View
8
Download
0
Category
Preview:
Citation preview
Preprint typeset in JHEP style - HYPER VERSION arXiv:1102.0440 [hep-th]
Towards a derivation of holographic
entanglement entropy
Horacio Casini,1 Marina Huerta1 and Robert C. Myers2
1Centro Atomico Bariloche and Instituto Balseiro
8400-S.C. de Bariloche, Rıo Negro, Argentina2Perimeter Institute for Theoretical Physics
Waterloo, Ontario N2L 2Y5, Canada
Abstract: We provide a derivation of holographic entanglement entropy for spherical
entangling surfaces. Our construction relies on conformally mapping the boundary
CFT to a hyperbolic geometry and observing that the vacuum state is mapped to
a thermal state in the latter geometry. Hence the conformal transformation maps
the entanglement entropy to the thermodynamic entropy of this thermal state. The
AdS/CFT dictionary allows us to calculate this thermodynamic entropy as the horizon
entropy of a certain topological black hole. In even dimensions, we also demonstrate
that the universal contribution to the entanglement entropy is given by A-type trace
anomaly for any CFT, without reference to holography.
Keywords: entanglement entropy, conformal anomaly, holography.arX
iv:1
102.
0440
v2 [
hep-
th]
27
Feb
2011
Contents
1. Introduction 1
2. The CFT story 4
2.1 Entanglement entropy in flat space 4
2.2 Thermal behaviour in R×Hd−1 10
2.3 Entanglement entropy in a cylindrical background 13
3. The AdS story 15
3.1 Entanglement entropy in flat space 18
3.2 Entanglement entropy in a cylindrical background 23
4. A CFT calculation 25
4.1 Mapping to de Sitter space 26
4.2 Thermodynamic entropy 27
4.3 Hyperbolic mapping 29
5. Discussion 31
1. Introduction
Entanglement entropy has become an important quantity for the study of quantum
matter. It allows one to distinguish new topological phases and characterize critical
points, e.g., [1, 2, 3]. Entanglement entropy has also been considered in discussions
of holographic descriptions of quantum gravity, in particular, for the AdS/CFT corre-
spondence [4, 5]. In this context, as well as characterizing new properties of holographic
field theories, e.g., [6], it has been suggested that entanglement entropy may provide
new insights into the quantum structure of spacetime [7].
The proposal [4, 5] of how to calculate holographic entanglement entropy is both
simple and elegant. In the boundary field theory, one would begin by chosing a par-
ticular spatial region V . The entanglement entropy between V and its complement
V would then be the von Neumann entropy of the density matrix which results upon
integrating out the field theory degrees of freedom in V . In the holographic calculation
– 1 –
which is proposed to yield the same entropy, one considers bulk surfaces m which are
”homologous” [8, 9] to the region V in the boundary (in particular ∂m = ∂V ). Then
one extremizes the area1 of m to calculate the entanglement entropy:
S(V ) =extm∼V
[A(m)
4GN
]. (1.1)
An implicit assumption in eq. (1.1) is that the bulk physics is described by (classical)
Einstein gravity. Hence we might note the similarity between this expression (1.1)
and that for black hole entropy. While this proposal to calculate the holographic
entanglement entropy passes a variety of consistency tests, e.g., see [8, 10, 11], there is
no concrete construction which allows one to derive this holographic formula (1.1).
A standard approach to calculating entanglement entropy in field theory makes use
of the replica trick [2, 12]. This approach begins by calculating the partition function
on an n-fold cover of the background geometry where a cut is introduced throughout
the exterior region V . If one were to apply this construction for the boundary CFT in
holographic framework, one would naturally produce a conical singularity in the bulk
geometry with an angular excess of 2π(n−1). However, without a full understanding of
string theory or quantum gravity in the AdS bulk, we do not understand how to resolve
the resulting conical curvature singularity and so it is not really possible work with this
bulk geometry in a controlled way, e.g., it is not possible to properly evaluate the
saddle-point action in the gravity theory. This issue was emphasized in [8] in critiquing
the attempted derivation of [9]. Further, it was demonstrated there that the approach
of [9] leads to incorrect results in holographic calculations of Renyi entropies.
Hence the replica trick does not seem a useful starting point in considering a deriva-
tion of holographic entanglement entropy. An interesting derivation of the holographic
entanglement entropy for a specific geometry was recently presented in [13, 14]. In
particular, the boundary CFT was placed on an R × Sd−1 background and the entan-
glement entropy was calculated for an entangling surface that divided the sphere into
two halves. It was argued that the holographic entanglement entropy was given by
the horizon entropy of a certain topological black hole whose horizon divided the bulk
geometry in half. In this paper, we clarify this construction and in doing so extend the
derivation to more general spherical entangling surfaces.
In section 2, we set aside holography and begin with a discussion of the entan-
glement entropy for spherical entangling surfaces for a general conformal field theory1If the calculation is done in a Minkowski signature background, the extremal area is only a saddle
point. However, if one first Wick rotates to Euclidean signature, the extremal surface will yield the
minimal area. In either case, the area must be suitably regulated to produce a finite answer. Further
note that for a d-dimensional boundary theory, the bulk has d+ 1 dimensions while the surface m has
d− 1 dimensions. We are using ‘area’ to denote the (d− 1)-dimensional volume of m.
– 2 –
(CFT). In particular, we use conformal transformations to map the causal development
of the interior region V to a new geometry which is the direct product of time with the
hyperbolic plane Hd−1. Further, we demonstrate that if the CFT began in the vacuum
in the original space, in this new geometry, we have a thermal bath whose temperature
is controlled by the size of the sphere. Hence the density matrix describing the CFT
inside the sphere becomes a thermal density matrix on the hyperbolic geometry (up to
a unitary transformation which, of course, preserves the entropy). Hence the entan-
glement entropy across the sphere becomes the thermodynamic entropy in the second
space. This construction was, in part, inspired by the appearance of a hyperbolic space
in the calculations of entanglement entropies across spheres for a free scalar in [15] and
it generalizes some older observations in ref. [16], again for free field theories.
While the results outlined above are quite general, in particular applying for any
arbitrary CFT, it may seem this discussion only replaces one difficult problem, the
calculation of the entanglement entropy, with another equally difficult problem, the
calculation of the thermal entropy in a hyperbolic space. However, we are particularly
interested in applying this result to the AdS/CFT correspondence. In this framework,
the standard holographic dictionary suggests that the thermal state in the boundary
CFT is dual to a black hole in the bulk gravity theory. Hence in section 3, we identify
the corresponding black hole dual to the thermal bath on the hyperbolic geometry.
The latter turns out to be a certain topological black hole with a hyperbolic horizon,
but also one which is simply a hyperbolic foliation of the empty AdS spacetime. Hence
with this bulk interpretation, we are able to compute the entanglement entropy as the
horizon entropy of the black hole. This construction is not limited to having Einstein
gravity in the bulk and so in fact, a general result is presented for any gravitational
theory.
A generalization of the replica trick [12] has been applied to relate entanglement
entropy to the trace anomaly for CFT’s in an even number of spacetime dimensions
[5, 17]. To be precise, with a general entangling surface, the universal coefficient of the
logarithmic contribution to the entanglement entropy is given by some linear combina-
tion of the central charges appearing in the trace anomaly for any CFT with even d.
The latter can be written as [18]
〈T µµ 〉 =∑
Bn In − 2 (−)d/2AEd . (1.2)
where Ed is the Euler density in d dimensions and In are the independent Weyl in-
variants of weight −d.2 A result of our holographic analysis in section 3 is that the
2Note that we have discarded a scheme dependent total derivative on the right-hand side of eq. (1.2).
For more details on our conventions here, see [14].
– 3 –
linear combination determining the universal contribution for a spherical entangling
surface reduces to be simply the coefficient of the A-type trace anomaly. In section
4, we obtain the same result for an arbitrary CFT, without reference to holography.
There, our approach again relies on using a conformal mapping of the geometry in the
entanglement entropy calculation. However, in this case, our conformal mapping takes
the CFT to (the static patch of) de Sitter space, where the state is again thermal, and
the entropy is again interpreted as ordinary thermodynamical entropy.
A connection of the coefficient of the logarithmic term in the entanglement entropy
of a sphere with the A-type anomaly was also found previously for d = 4 by [17].
Further, this connection was also noted for free scalar fields in any even dimension in
[19]. Let us note that there have been a number of other recent works related to the
entanglement entropy of free fields for a spherical entangling surface [15, 20, 21, 22].
2. The CFT story
In this section, we describe the entanglement entropy of any general conformal field the-
ory for spherical entangling surfaces in certain background geometries. We emphasize
that the discussion here is made purely in the context of quantum field theory (QFT),
without reference to holography. However, the results found here will set the stage
of a holographic calculation of the same entanglement entropy using the AdS/CFT
correspondence in section 3. In particular, we consider here a d-dimensional CFT in
Minkowski space R1,d−1 and examine the entanglement entropy across a spherical sur-
face Sd−2. As described above, we show that the causal development of the region inside
the Sd−2, which we denote D, can be mapped to a space H ≡ R ×Hd−1. Further we
show that the vacuum correlators in D are conformally mapped to thermal correlators
in the space H. Hence we are able to show that the density matrix on D is given as a
Gibbs state of a local operator which is just constructed by conformally mapping the
(curved space) Hamiltonian of the CFT in H back to D. With this construction, the
entanglement entropy across the sphere becomes the thermal entropy in H. In section
2.3, we demonstrate that the same results apply when we begin with the CFT in a
cylindrical background geometry, i.e., R× Sd−1.
2.1 Entanglement entropy in flat space
Consider an arbitrary quantum field theory in d-dimensional Minkowski space R1,d−1.
As shown in figure 1, we introduce a smooth entangling surface Σ which divides the
t = 0 time slice into two parts, a region V and its complement V . Upon integrating out
the degrees of freedom in V , we are left with the reduced density matrix ρ describing
– 4 –
Figure 1: The division of the t = 0 time slice into two regions V and V .
the remaining degrees of freedom in the region V . The entanglement entropy across Σ
is then just the von Neumann entropy of ρ, i.e.,
SΣ = −tr(ρ log ρ) . (2.1)
However, we note that the definition and interpretation of these objects in the contin-
uum QFT requires a regularization, as will become apparent below.
Since the reduced density matrix is both hermitian and positive semidefinite, it
can be expressed as
ρ = e−H (2.2)
for some hermitian operator H. In the literature on axiomatic quantum field theory, H
is known as the modular Hamiltonian [23].3 We emphasize that generically H is not a
local operator. That is, it can not be represented as some local expression constructed
with the fields on V . However, the modular Hamiltonian still plays an important role
since the unitary operator U(s) = ρis = e−iHs generates a symmetry of the system.
One easily sees that
tr(ρU(s)OU(−s)) = tr(ρO) , (2.3)
3The same operator, referred to as the ‘entanglement Hamiltonian’, has recently also appeared in
studies of the ‘entanglement spectrum’ of topological phases of matter [24].
– 5 –
for any operatorO localized inside V . In fact, because of causality, this symmetry group
transforms the algebra of operators inside the causal development4 of V into itself. In
the algebraic approach to QFT, this one-parameter group of transformations U(s) is
called the modular group [23]. Further if these transformations are extended to com-
plex parameters, one finds that correlators obey the KMS (Kubo-Martin-Schwinger)
periodicity relation in imaginary time. Defining O(s) = U(s)OU(−s), this relation is
easily established with
tr(ρO1(i)O2) = tr(ρU(i)O1U(−i)O2) = tr(ρO2O1) . (2.4)
The last equality follows using U(i) = ρ−1, U(−i) = ρ and the cyclicity of the trace.
Hence formally we can say the state ρ is thermal with respect to the time evolution
dictated by the internal symmetry U(s), with an inverse temperature β = 1. However,
we emphasize that these are formal expressions. As noted above, generically H is not
local and U(s) does not generate a local (geometric) flow on D. For example, if we
begin with a local operator defined at a point, O = φ(x), then generally the operator
O(s) will no longer have this simple form of being defined at a point.
However, there are special cases where the modular flow and the modular Hamil-
tonian are in fact local. One well-known example is given by Rindler space R, i.e., the
wedge of Minkwoski space corresponding to the causal development of the half space
X1 > 0. In this case for any QFT, the modular Hamiltonian is just the boost generator
in the X1 direction. This result is commonly known as the Bisognano-Wichmann the-
orem [25]. In this case then, the modular transformations act geometrically in Rindler
space. They map the algebra of operators A(B) localized in a region B ⊆ R to the
algebra of operators A(Bs) in the region Bs,
U(s)A(B)U(−s) = A(Bs) , (2.5)
where Bs is the mapping of B by the boost transformation. More explicitly, the modular
flow is given by
X±(s) = X±e±2πs , (2.6)
where X± ≡ X1±X0 are the null coordinates with 0 ≤ X± < +∞ in R. Of course, the
modular flow leaves all other coordinates invariant, i.e., X i(s) = X i for i = 2, ..., d− 1.
Interpreted in the sense of Unruh [26], the state in R is thermal with respect to the
notion of time translations along the boost orbits. If we choose conventional Rindler
coordinates,
X±(s) = z e±τ/R , (2.7)
4The causal development of V , which we denote D, is the set of all points p for which all causal
curves through p necessarily intersect V .
– 6 –
the Minkowski space metric becomes
ds2 = dX+ dX− +d−1∑i=2
(dX i
)2= − z
2
R2dτ 2 + dz2 +
d−1∑i=2
(dX i
)2. (2.8)
We introduced an arbitrary scale R above in eq. (2.7) to ensure that τ has the standard
dimensions of length. With this choice, the Rindler state is thermal with respect to Hτ ,
the Hamiltonian generating τ translations, with a temperature T = 1/(2πR). Hence
the density matrix can be simply written as the thermal density matrix
ρR =e−βHτ
Zwhere Z = tr
(e−βHτ
). (2.9)
With this notation, the modular flow (2.6) on R simply corresponds to the time trans-
lation
τ → τ + 2πR s (2.10)
and the modular Hamiltonian HR is simply related to Hτ with HR = 2πRHτ + logZ.
In the following, we are particularly interested in the entanglement entropy for the
case where the entangling surface Σ is a (d–2)-dimensional sphere of radius R. Hence
the region V becomes the ball bounded by this Sd−2. Let us define the null coordinates
x± ≡ r± t with t = x0 and the radial coordinate r =√
(x1)2 + ...+ (x(d−1))2. Then the
causal development D of the ball is the spacetime region defined by x+ ≤ R ∩ x− ≤R — implicitly, we are assuming x+ + x− = 2r ≥ 0 in this definition. Further we
wish to consider the special case where the entanglement entropy is calculated in this
geometry for a conformal field theory. In this case, the modular Hamiltonian on D will
in fact be a local operator.
This last fact can be derived making use of the previous result for the Rindler
wedge R. To begin, we observe that there is a special conformal transformation (and
translation) which maps the Rindler wedge to the causal development D of the ball
(e.g., see [23, 27])
xµ =Xµ − (X ·X)Cµ
1− 2(X · C) + (X ·X)(C · C)+ 2R2Cµ , (2.11)
with Cµ = (0, 1/(2R), 0, ...0). It is straightforward to show that X± ≥ 0 covers x± ≤ R.
Explicitly, eq. (2.11) yields ds2 = ηµνdXµdXν = Ω2 ηµνdx
µdxν where the conformal
pre-factor can be written as
Ω = 1− 2(X · C) + (X ·X)(C · C) .
= (1 + 2(x · C) + (x · x)(C · C))−1 . (2.12)
– 7 –
Further eq. (2.11) maps the flow (2.6) to the following geometric flow in D,
x±(s) = R(R + x±)− e∓2πs(R− x±)
(R + x±) + e∓2πs(R− x±). (2.13)
Now it is not difficult to show that the induced flow (2.13) gives the modular flow
of the CFT on D. Recall that there is a unitary operator, which we denote U0, in
the CFT which implements the conformal transformation associated eq. (2.11). For
example, the primary operators of the CFT transform locally as (considering spinless
operators for simplicity)
φ(x) = Ω(X)∆ U0 φ(X)U−10 (2.14)
where ∆ is the scaling dimension of the field. Since this mapping is a conformal
symmetry of Minkowski space, it leaves the vacuum state invariant in a conformal
theory, i.e., U0|0〉 = |0〉. At a more practical level, vacuum correlators on R are
mapped to vacuum correlators on D
〈φ1(x1) · · ·φn(xn)〉 = Ω(X1)∆1 · · ·Ω(Xn)∆n 〈φ1(X1) · · ·φn(Xn)〉 . (2.15)
Now we may apply U0 to construct the quantum operator generating the modular flow
on D with
UD(s) = U0 UR(s)U−10 . (2.16)
One can show that UD(s) generates the ‘modular transformations’ for the sphere Σ
[27]. To confirm that eq. (2.16) yields the modular operator, we must verify that
two conditions are satisfied [23]: i) the transformation must be a symmetry of the
correlators, as in eq. (2.3), and ii) the correlators must obey the KMS periodicity, as
in eq. (2.4). The first condition eq. (2.3) is evident, since (2.16) is just a particular
conformal symmetry of the theory which keeps the sphere invariant.
Turning to the second condition, we note that the correlator on the right hand side
of (2.15) satisfies the KMS condition with respect to the flow (2.6). Further combining
eqs. (2.6) and (2.12), it is also clear that each of the pre-factors involving Ω is also
periodic in s with imaginary period i. Hence, the correlator on the left hand side also
satisfies the KMS condition but with respect to the induced flow (2.13). Hence given
that conditions (i) and (ii) are both satisfied, we conclude that (2.16) are indeed the
modular transformations and the geometric flow (2.13) is the modular flow on D.
Acting on a (spinless) primary operator in the CFT, one finds
UD(s)φ(x[s0])UD(−s) = Ω(x[s0])∆ Ω(x[s0 + s])−∆ φ(x[s0 + s]) , (2.17)
using eq. (2.14). Here, our notation x[s0] indicates that the position of the operator
flows according to eq. (2.13). Note then that UD(s) acts both to translate the operator
– 8 –
along the geometric flow (2.13) and multiplies it with a particular (c-number) pre-
factor. The key point, however, is that UD(s) remains a local transformation, taking
an operator defined at a point to that same operator defined at a new point.
In fact, we can produce an explicit expression for the modular Hamiltonian at this
point. The construction is simplest if we focus on V , the ball of radius R, in the time
slice x0 = 0. Since the modular Hamiltonian is the generator for the transformation
(2.17), it will produce an infinitesimal shift δs on the surface x0 = 0:
δx0 = 2πR2 − r2
2Rδs and δr = 0 . (2.18)
At the same time, an infinitesimal shift δs produces a pre-factor in eq. (2.17) which
takes the form
Ω(x[s0])∆ Ω(x[s0 + δs])−∆∣∣x0=0' 1−∆
∂sΩ
Ω
∣∣∣∣x0=0
δs . (2.19)
However, using eqs. (2.12) and (2.18), it is straightforward to show that ∂sΩ = 0 when
evaluated on the x0 = 0 surface and so the pre-factor is simply 1 at this order. Hence
the modular Hamiltonian simply induces the infinitesimal flow (2.18) away from x0 = 0
and we can identify the corresponding operator in the CFT as
HD = 2π
∫V
dd−1x(R2 − r2)
2RT 00(x) + c′ , (2.20)
where T µν is the conformal traceless stress tensor and c′ is some constant added to
ensure that the corresponding density matrix is normalized with unit trace. Hence this
expression explicitly shows the modular Hamiltonian as a local operator, for a CFT in
the causal development D.
One point that we feel is worth emphasizing is that the standard CFT correlators
in D transform ‘covariantly’ under UD(s) — that is, they are not invariant, as one
might naıvely surmise from eq. (2.3). This is simply the observation that if we begin
with O = φ1(x1[s0]) · · ·φn(xn[s0]) in eq. (2.3), then eq. (2.17) dictates that generally
O(s) 6= φ1(x1[s0 + s]) · · ·φn(xn[s0 + s]). Instead, the correlators of (spinless) primary
operators transform as
Ω(x1[s0])−∆1 · · ·Ω(xn[s0])−∆n 〈φ1(x1[s0]) · · ·φn(xn[s0])〉 (2.21)
= Ω(x1[s0 + s])−∆1 · · ·Ω(xn[s0 + s])−∆n 〈φ1(x1[s0 + s]) · · ·φn(xn[s0 + s])〉 .
Clearly, the idea of ‘transplanting’ the modular flow with a conformal transforma-
tion can be used to obtain the modular Hamiltonian (and the density matrix) for a CFT
– 9 –
in other conformally connected geometries. Note that the state in the transformed space
has to be chosen as the conformally transformed state, which is the one that makes
(2.21) work. In the present case of a mapping between the Rindler wedge R and the
causal development D, the transformation is a conformal symmetry of Minkowski space
and the transformed state is again the Minkowski vacuum. Of course, for operators
inside D, the vacuum coincides with the density matrix ρ, meaning that they both give
the same expectation values 〈0|O|0〉 = tr(ρO).
2.2 Thermal behaviour in R×Hd−1
Next we would like to extend this approach of transplanting modular flows to relate
the density matrix of a CFT on D to that on a new geometry R × Hd−1, which we
will denote as H in the following. In particular, we will show that beginning with the
Minkowski vacuum for an arbitrary CFT, the density matrix on D becomes a thermal
density matrix on H. In fact, our result is a generalization of previous observations
made in ref. [16]. There the generation of a thermal state by conformal mappings was
observed for free conformal field theories in d = 4.
One indication that the claim above holds comes from first returning to the Rindler
wedge (2.8). Here, we can write the metric as
ds2 = Ω2
(−dτ 2 +
R2
z2
[dz2 +
d−1∑i=2
(dX i
)2
]), (2.22)
where Ω = z/R. Hence with a conformal transformation which eliminates the pre-
factor Ω2, the Rindler metric is mapped precisely to the metric on R × Hd−1. Note
that the scale R now sets the curvature scale of the hyperbolic plane Hd−1. As in the
previous section, there is a unitary operator which maps the CFT from R to H, which
we will denote U1. Further we may again apply U1 to construct the operator generating
the modular flow on H with
UH(s) = U1 UR(s)U−11 . (2.23)
In this particular case, the conformal mapping is time (i.e., τ) independent and so U1
acts ‘trivially’ on the modular Hamiltonian. That is, HH = 2πRHτ + logZ where
Hτ is now the generator of τ -translations in H. Therefore the new density matrix ρHinherits the same thermal character as in eq. (2.9). Hence combining this result with
those in the previous section, we can map the reduced density matrix on D to a thermal
density matrix on R and then to a thermal density matrix on H. However, let us step
back and establish the relationship between ρD and ρH directly.
– 10 –
First, we present the conformal transformation which maps D to H. We start with
the flat space metric in polar coordinates:
ds2 = −dt2 + dr2 + r2 dΩ2d−2 , (2.24)
where dΩ2d−2 is the line element on a round Sd−2 with unit curvature. Our entangling
surface Σ is again the sphere r = R on the surface t = 0. Now we make the coordinate
transformation
t = Rsinh(τ/R)
coshu+ cosh(τ/R), (2.25)
r = Rsinhu
coshu+ cosh(τ/R).
One can readily verify the above metric (2.24) becomes
ds2 = Ω2[−dτ 2 +R2
(du2 + sinh2u dΩ2
d−2
)]where Ω = (coshu+ cosh(τ/R))−1 . (2.26)
After the conformal transformation which eliminates the pre-factor Ω2, we again recog-
nize the resulting line element as the metric on R×Hd−1. The curvature on the latter
hyperbolic space is
Rijk` = − 1
R2
(δikδj
` − δi`δjk). (2.27)
We note that fixing the curvature scale to match R, the radius of the sphere Σ, is an
arbitrary but convenient choice. Finally we observe that
τ → ±∞ : (t, r)→ (±R, 0)
u→∞ : (t, r)→ (0, R)
Hence the new coordinates cover preciselyD, the causal development of the region inside
of Σ. Hence we have our conformal mapping from D to H. We are again considering an
arbitrary CFT and so there is a unitary transformation U2 implementing the conformal
transformation (2.25) on the Hilbert space of the CFT.5
Now, we wish to relate the two reduced density matrices: ρD describing the vacuum
state on D and ρH for the corresponding state on H. In particular, we wish to establish
that the latter corresponds to a thermal density matrix with temperature T = 1/(2πR).
For the latter to hold, we must verify two conditions: First, the modular flow in H must
correspond to ordinary time translations and these translations must be a symmetry of
5Of course, this operator is related to those appearing in the previous discussion by U2 = U1U−10 .
– 11 –
the correlators. Second, the correlators on H must be periodic for an imaginary shift
of τ by 2πR.
As a first step, we use eq. (2.25) to write
x± = R1− e−v±
1 + e−v±(2.28)
where we have defined v± ≡ u±(τ/R) — recall from above that x± = r±t. From these
expressions, it is straightforward to show that the modular flow (2.13) on D corresponds
to a time translation
τ → τ + 2πR s (2.29)
in H — just as in eq. (2.10). Hence as desired, the modular Hamiltonian induces a
flow along time translations. However, for a thermal state, this flow must correspond
to a symmetry of the correlators, i.e., it is not enough that the correlators transform
covariantly under the modular flow. The primary fields transform by the analog of
eq. (2.14) replacing U0 by U2 and hence we can write the relation for the correlation
functions under the conformal transformation as
〈φ′1(x′1) · · ·φ′n(x′n)〉H = Ω(x′1)∆1 · · ·Ω(x′n)∆n 〈φ1(x1) · · ·φn(xn)〉D . (2.30)
where Ω(x′) is the conformal pre-factor in eq. (2.26)). Now the factors Ω in (2.30)
are not invariant under shifts of τ . However, we also noted above in eq. (2.21) that
the modular flow on D transforms the correlators on the right-hand side non-trivially.
Having the connection (2.29) between translations in τ and s, a simple calculation
shows the τ dependence of the pre-factors precisely cancels that of the correlators on
D. Hence the conformally mapped correlators on H are in fact invariant under shifts in
τ . Hence the modular flow inH is just the ordinary time translations (2.29). Examining
eq. (2.30) further, we find the second requirement, that the correlators on H must be
periodic for an imaginary shift of τ by 2πR also follows. First, given eq. (2.26), it is
clear that each of conformal pre-factors satisfies this condition. Next the correlators on
D satisfy the KMS condition (2.4) and from eq. (2.29), it follows that the imaginary
shift of s by i corresponds to the desired shift of τ .
Hence the two requirements above are both satisfied and so we have established
the desired result: The conformal mapping (2.26) takes the CFT in the Minkowski
vacuum on D to a thermal state on H where the latter is thermal with respect to the
standard Hamiltonian Hτ with a physical temperature T = 1/(2πR). That is, we have
ρH = e−βHτ/Z. Now the unitary operator U2 will map between the reduced density
matrix on D and the thermal density matrix on H. More explicitly, we may write the
density matrix on D as
ρD ≡ e−HD =1
ZU−1
2 e−βHτU2 . (2.31)
– 12 –
Since the von Newman entropy is invariant under unitary transformations, the entan-
glement entropy across the sphere Σ in flat space is then equal to the corresponding
thermal entropy of the Gibbs state in H.
However, some care must be exercised for the equality of the entropies to hold since
both of these quantities are divergent. In particular, in order to maintain the equality,
we must also use the conformal transformation to map between the cut-off procedures
in two spaces. In the case of entanglement entropy (2.1), there is a UV divergence at
the boundary Σ and so we need to introduce a short distance cut-off scale δ. That is, we
only take contributions down to r = R− δ where δ/R 1.6 In the case of the thermal
entropy, there is an IR divergence because we have a uniform entropy density but the
volume of the spatial slices, i.e., Hd−1, is infinite. Hence we regulate the entropy by
integrating out to some maximum radius, u = umax where umax 1. Now, consistency
demands that the cut-off’s should be related by the conformal mapping between the
two spaces. If we focus on the t = 0 slice (or equivalently the τ = 0 slice), then
eq. (2.25) yields
R− δ = Rsinhumax
coshumax + 1. (2.32)
After a bit of algebra, we find
exp(−umax) =δ/R
2− δ/R' δ
2R. (2.33)
It is interesting that here the conformal mapping has introduced a UV/IR relation be-
tween the relevant cut-offs in the two CFT states. Of course, similar relations commonly
occur in the AdS/CFT correspondence but here we have a purely CFT calculation.
2.3 Entanglement entropy in a cylindrical background
In this section, we develop the analogous account of the ‘thermalization by conformal
mapping’ found above for the case where the original background geometry is R×Sd−1.
That is, we wish to calculate the entanglement entropy across a sphere of a fixed angular
size embedded in the t = 0 slice of the static Einstein universe. As before, we map
the causal development of the interior of this sphere to R × Hd−1 with a conformal
transformation and show that the resulting density matrix describes a thermal state.
Hence the entanglement entropy becomes the thermal entropy of the Gibbs state in H.
Below we only present the salient features of the proof as conceptually it is completely
analogous to the discussion above.
6A regularization of the entanglement entropy implementing a distance cutoff to the entangling
sphere can be made precise using the mutual information [28].
– 13 –
We start with the metric in d-dimensional cylindrical spacetime with topology
R× Sd−1:
ds2 = −dt2 +R2(dθ2 + sin2θ dΩ2
d−2
). (2.34)
The entangling surface Σ will now be the sphere θ = θ0 on the surface t = 0. Now
make the coordinate transformation
tan(t/R) =sin θ0 sinh(τ/R)
coshu+ cos θ0 cosh(τ/R), (2.35)
tan θ =sin θ0 sinhu
cos θ0 coshu+ cosh(τ/R).
One can readily verify the above metric (2.34) becomes
ds2 = Ω2[−dτ 2 +R2
(du2 + sinh2u dΩ2
d−2
)](2.36)
where Ω2 =sin θ2
0
(coshu+ cos θ0 cosh(τ/R))2 + sin θ20 sinh2u
.
Hence, after eliminating the conformal factor Ω2 in the first line, we again recognize
the final line element as the metric on R×Hd−1,
ds′2 = −dτ 2 +R2(du2 + sinh2u dΩ2
d−2
), (2.37)
precisely as in eq. (2.26). The curvature scale of the hyperbolic space is R, as in
eq. (2.27). As before, we note that this is an arbitrary but convenient choice for the
curvature. In this case, R also corresponds to the radius of curvature of the Sd−1 in
the Einstein universe, rather than the size of the entangling sphere Σ, as we will see
below.
Examining what portion of the original Einstein universe (2.34) is covered by the
new coordinates, we find
τ → ±∞ : (t, θ)→ (±Rθ0, 0)
u→∞ : (t, θ)→ (0, θ0)
Hence, the new coordinates cover precisely the causal development D of the ball en-
closed by the entangling sphere Σ. As in section 2.3, examining the effect of the con-
formal mapping on CFT correlators, we find that the vacuum correlators in D become
thermal correlators on H with temperature T = 1/(2πR).
As before, the entanglement entropy across the sphere Σ in the Einstein space
matches the thermal entropy of the Gibbs state in H but we must be careful in regulat-
ing the two expressions. In the case of entanglement entropy, there is a UV divergence
– 14 –
at the surface Σ and we need to introduce a short distance cut-off. The natural cut-off
is a minimal angular size δθ ( 1) and only taking contributions out to θ = θ0−δθ. As
before, the thermal entropy has an IR divergence because we have a uniform entropy
density in the the infinite volume of the spatial Hd−1 slices. We again regulate this en-
tropy by only integrating out to u = umax where umax 1. Now consistency demands
that the cut-off’s should be related by the conformal mapping between D and H. If we
focus on the t = 0 slice (or equivalently the τ = 0 slice), then eq. (2.35) yields
tan(θ0 − δθ) =sin θ0 sinhumax
cos θ0 coshumax + 1. (2.38)
With a bit of algebra, we find
exp(−umax) =1
2 sin θ0
tan δθ + tan θ0(sec δθ − 1)
1− cot 2θ0 tan δθ' δθ
2 sin θ0
. (2.39)
Hence as in the previous section, we have an interesting UV/IR relation between the
cut-offs in the two CFT calculations.
3. The AdS story
In the previous section, we related the problem of calculating entanglement entropy
for a spherical entangling surface to calculating the thermal entropy of a Gibbs state
in R×Hd−1, where the temperature is of the same order as the curvature scale of the
hyperbolic geometry. While this is quite general result that applies for any arbitrary
CFT, it seems that we have only replaced one difficult problem with another equally
difficult problem. However, we are particularly interested in applying this result to
the AdS/CFT correspondence. In this framework, the standard holographic dictionary
suggests that the thermal state in the boundary CFT is dual to a black hole in the
bulk gravity theory. Hence if we are able to identify the corresponding black hole, the
thermal entropy of the CFT can be calculated as the horizon entropy of the black hole.
Precisely this problem was encountered in [13, 14] in a related calculation of the en-
tanglement entropy for a specific geometry. In this case, the boundary CFT was placed
in a cylindrical space, R × Sd−1 — that is, the (complete) boundary of AdSd+1. The
problem was then to determine the entanglement entropy when the (d–1)-dimensional
sphere was divided in half along the equator, i.e., the entangling surface was chosen as
a maximal Sd−2 in a constant time slice. In [13, 14], it was argued that the entangle-
ment entropy could be identified with the horizon entropy of a topological AdS black
hole, for which the large radius limit (at fixed time) covered precisely one half of the
boundary Sd−1. In fact, the topological black hole was simply an R×Hd−1 foliation of
the AdSd+1 geometry.
– 15 –
Given the CFT discussion of section 2, we now have a clearer understanding of
the calculations in [13, 14] and in turn, the calculations there suggest the necessary
approach to implement our results from the previous section in a holographic setting.
In particular, we begin with the boundary CFT in its vacuum either in Minkowski space
R1,d−1 or the cylindrical space R×Sd−1. The dual gravity description is simply the pure
AdSd+1 geometry in a coordinate system which foliates the spacetime with R1,d−1 or
R×Sd−1 surfaces. To calculate the entanglement entropy of the CFT across some sphere
Σ in the boundary geometry, we must next find a new foliation of the AdSd+1 in terms
of R×Hd−1. This foliation is chosen to cover the ball enclosed by Σ in the asymptotic
boundary geometry. In fact, the causal development D of this ball will be covered in
the asymptotic limit of the R × Hd−1 foliation. Implicitly, we will have implemented
the conformal mapping of the causal development D of the ball to R × Hd−1 in the
boundary CFT with the bulk coordinate transformation between the two foliations of
the AdS space. As noted in [13, 14], the new hyperbolic foliation can be interpreted
in terms of a topological black hole [29, 30] and so on the R ×Hd−1 background, the
boundary CFT is naturally seen to be in a thermal state. The temperature of the latter
thermal state is given by the Hawking temperature of the black hole horizon and we
will find that we precisely reproduce the result of section 2: T = 1/(2πR). Further,
since as discussed in section 2, the entanglement entropy across the sphere Σ can be
calculated as the thermal entropy of the Gibbs state in the R × Hd−1 geometry, the
same entropy is given by the horizon entropy of the bulk black hole in the holographic
setting.
Let us emphasize that this approach gives us a derivation of the entanglement
entropy of the holographic CFT, in the case of a spherical entangling surface. Given
the insights of the previous section, we are simply applying the standard AdS/CFT
dictionary to implement the results there in a holographic framework. Further let us
note that if the bulk theory was Einstein gravity, then our results of the entanglement
entropy for this particular set of geometries would precisely match the holographic en-
tanglement entropy calculated with the extremal area prescription (1.1) conjectured by
Ryu and Takayanagi [4, 5]. In this regard, our results provide a nontrivial confirmation
of their proposal.
In fact, since we have a derivation of the entanglement entropy, we do not need to
limit our discussion to Einstein gravity. Hence in the following discussion, we allow the
bulk gravity to be described by any arbitrary covariant action of the form
I =
∫dd+1x
√−gL(gab, Rab
cd,∇eRabcd, · · · ,matter) . (3.1)
Hence one might apply our analysis in the context of the low energy effective action
– 16 –
in string theory, where the contributions of higher curvature terms are controlled to
make small corrections to the results of the Einstein theory [31]. Alternatively, our
results would also be applicable to a situation where higher curvature contributions are
finite, as in the recent studies of the AdS/CFT correspondence [32] with Lovelock [33]
or quasi-topological [34] gravity. In the following, we will assume that the couplings of
the above theory are chosen so that the theory has a vacuum solution which is AdSd+1
spacetime with a curvature scale L. Implicitly, we will also assume that the bulk
couplings are further constrained so that the boundary CFT has physically reasonable
properties (e.g., it should be causal and unitary) — for example, see [14, 34, 32].
An essential step in the following will be calculating the horizon entropy of the
bulk black hole. In general, the horizon entropy can be calculated using Wald’s entropy
formula [35]
S = −2π
∫horizon
dd−1x√h
∂L∂Rab
cd
εab εcd , (3.2)
which can be applied for any (covariant) action, as assumed above in eq. (3.1). Note
that εab denotes the binormal to the horizon. Of course, as described above, the case
of interest is a topological black hole which in fact corresponds to a R×Hd−1 foliation
of AdSd+1. In this case, the integrand in eq. (3.2) is constant across the horizon and so
the total entropy diverges. Of course, this divergence is expected given the discussion
towards the end of section 2.2 and we will return to this point in the following. However,
at this point, we use some results from [14] to re-express the integrand as
δLδRab
cd
εab εcd
∣∣∣∣AdS
= −Γ(d/2)
πd/2a∗dLd−1
(3.3)
where L is the AdS curvature scale. The (dimensionless) constant a∗d is a central charge
that characterizes the number of degrees of freedom in the boundary CFT [14]. In the
case where d, the dimension of boundary theory, is even, a∗d is precisely equal to A,
the coefficient of the A-type trace anomaly (1.2) in the CFT [14]. We stress that
eq. (3.3) relies on the fact that the background geometry in which we are evaluating
this expression is simply the AdSd+1 spacetime. Substituting this result into eq. (3.2)
leaves us with
S =2 Γ(d/2)
πd/2−1
a∗dLd−1
∫horizon
dd−1x√h . (3.4)
We now turn to the detailed determination of the hyperbolic foliations discussed
above. We begin by examining the entangling surface being a sphere in flat space in
the next section and then consider the case with a cylindrical background in section
3.2.
– 17 –
3.1 Entanglement entropy in flat space
Here we want to present a holographic calculation that implements the discussion of
entanglement entropy of a CFT in flat space given in section 2.1. So let us begin with
a standard description of the AdSd+1 geometry as the following surface
−y2−1 − y2
0 + y21 + · · ·+ y2
d = −L2 (3.5)
embedded in R2,d with
ds2 = −dy2−1 − dy2
0 + dy21 + · · ·+ dy2
d . (3.6)
Now we can connect this geometry to the standard Poincare coordinates on AdSd+1
space with
y−1 + yd =L2
z, ya =
L
zxa with a = 0, · · · , d− 1 . (3.7)
Now eq. (3.5) becomes a constraint which yields
y−1 − yd = z +1
zηab x
axb . (3.8)
The induced metric on the hyperboloid (3.5) then becomes the standard AdSd+1 metric
ds2 =L2
z2
(dz2 + ηab dx
adxb)
(3.9)
where the bulk spacetime is foliated with slices corresponding to flat d-dimensional
Minkowski space. As usual in the AdS/CFT framework, we take the asymptotic limit
z → 0 and remove a factor of L2/z2 from the boundary metric. This yields ds2CFT =
ηab dxadxb as the metric in which the dual CFT lives when making our holographic
calculations.
Another useful foliation of AdSd+1 is given by
y−1 = ρ coshu , y0 = ρ sinh(τ /L) , yd = ρ cosh(τ /L) , (3.10)
y1 = ρ sinhu cosφ1 , y2 = ρ sinhu sinφ1 cosφ2 , · · · yd−1 = ρ sinhu sinφ1 sinφ2 · · · sinφd−2 .
In this case, the constraint imposed by eq. (3.5) yields ρ2 = ρ2 − L2 and the induced
metric on the hyperboloid (3.5) becomes
ds2 =dρ2
ρ2
L2 − 1−(ρ2
L2− 1
)dτ 2 + ρ2
(du2 + sinh2 u dΩ2
d−2
)(3.11)
where dΩ2d−2 = dφ2
1 + sin2 φ1
(dφ2
2 + sin2 φ2 (dφ23 + · · · )
)is the line element on a unit
Sd−2. Of course, the bracketed expression in eq. (3.11), which is multiplied by ρ2,
– 18 –
corresponds to the line element on a (d–1)-dimensional hyperbolic plane Hd−1 with
unit curvature. Hence as desired, we are foliating the AdSd+1 space with surfaces with
a R × Hd−1 geometry in the above metric (3.11). Taking the asymptotic limit and
removing a factor of ρ2/L2, we find the boundary metric is precisely that given in
eq. (2.37) with R = L and τ = τ .
For our purposes, the essential feature of the second metric (3.11) is that it can be
interpreted as a topological black hole [29, 30] where the horizon at ρ = L has uniform
negative curvature. We would like to see what portion of the asymptotic AdS boundary
in the Poincare coordinates is covered by the exterior of this topological black hole. We
begin by considering the bifurcation surface (i.e., the intersection of the past and future
horizons) in eq. (3.11) which corresponds to ρ = L and any finite value of τ . Combining
eqs. (3.7) and (3.10), we find on this surface
y−1 + yd =L2
z= L coshu , y2
1 + · · ·+ y2d−1 =
L2
z2r2 = L2 sinh2 u . (3.12)
where we have introduced a radial coordinate for the boundary coordinates, i.e., r2 =
(x1)2 + ... + (xd−1)2. Hence we can see that the bifurcation surface intersects the
asymptotic boundary in the Poincare coordinates with
r2 = z2 sinh2 u = L2 tanh2 u →u→∞
L2 . (3.13)
That is, on a sphere of radius r = L in the Minkowski coordinates of the boundary CFT.
This will be the entangling sphere Σ in the calculation of the entanglement entropy.
One can work harder to find the precise connection between the two coordinate systems
on the boundary of AdSd+1. The resulting coordinate transformation is precisely that
given in eq. (2.25) with R = L and τ = τ .7
Next we must determine the Hawking temperature of our topological black hole
(3.11). A straightforward calculation shows that T = 1/(2πL). Given that the entan-
gling sphere has a radius R = L, this result precisely matches the temperature found in
section 2, where the discussion applied for any CFT without reference to holography.
Hence as discussed in the introductory remarks of this section, changing coordinates
in the bulk between the Poincare and hyperbolic foliations implements the conformal
mapping from D, the causal development of the sphere at r = L, to R × Hd−1. As
expected from the general considerations of section 2, this mapping generates a thermal
state in the R×Hd−1 geometry. The entanglement entropy across the sphere r = L in
flat space equals the thermal entropy of this Gibbs state and further, in the holographic
setting, the latter is given by the horizon entropy of the bulk black hole.
7In fact, examining the asymptotic relation between the Poincare and hyperbolic foliations is how
we originally derived the transformation (2.25).
– 19 –
We leave the calculation of the entropy
Figure 2: A slice of constant t through the
AdSd+1 spacetime. The solid (red and green)
arcs each represent codimension two surfaces
with constant curvature −1/L2. These are
all connected by isometries of the AdSd+1 ge-
ometry. The red arcs are connected by a
boost, as described in the main text. The
dashed blue line represents a regulator sur-
face at large radius.
for a bit later. First we would like to ex-
tend the previous holographic calculations
to a sphere in Minkowski space with an
arbitrary radius R, as in eq. (2.25). Es-
sentially we want to move the bifurcation
surface above to a new position that in-
tersects the boundary in a different way,
as illustrated in figure 2. An interesting
observation is that in fact all of the sur-
faces illustrated in this figure are ‘identi-
cal’ in that they are connected by an isom-
etry of the AdSd+1 geometry. Hence one
finds that the bulk metric for the hyper-
bolic foliation in which any of these sur-
faces appears as the bifurcation surface is
identical to that in eq. (3.11). Given these
isometries, one might ask how the physics
(e.g., the entropy) can change. However,
in the AdS/CFT framework, calculations
are always made with reference to a regu-
lator surface in the asymptotic region, e.g.,
z = zmin in the Poincare metric (3.9). Be-
low, we will see that the regulator surface plays an essential role in the calculation of the
entanglement entropy. This surface is typically not invariant under the isometries that
connect the various candidate bifurcation surfaces and consequently, e.g., the entropy
will be different with different choices.
One simple isometry relating different bifurcation surfaces, illustrated in figure
2, is a boost in the embedding space of the yA coordinates. In particular, a boost
in the (y−1, yd)-plane leaves the hyperboloid (3.5) invariant while transforming the
coordinates:
y′−1 = cosh β y−1 − sinh β yd , (3.14)
y′d = cosh β yd − sinh β y−1 .
Now we make the trivial substitution to replace eq. (3.10) with
y′−1 = ρ coshu , y0 = ρ sinh(τ /L) , y′d = ρ cosh(τ /L) , (3.15)
y1 = ρ sinhu cosφ1 , y2 = ρ sinhu sinφ1 cosφ2 , · · · yd−1 = ρ sinhu sinφ1 sinφ2 · · · sinφd−2 .
– 20 –
As before, the constraint (3.5) yields ρ2 = ρ2 − L2 and the induced metric is identical
to that given in eq. (3.11). We leave our choice of Poincare coordinates unchanged as
in eq. (3.7). Hence to relate the two coordinate patches, it is useful to write
y−1 = cosh β ρ coshu+ sinh β√ρ2 − L2 cosh(τ /L) , (3.16)
yd = cosh β√ρ2 − L2 cosh(τ /L) + sinh β ρ coshu .
Again, this hyperbolic foliation (3.11) lends itself to an interpretation as a topo-
logical black hole. However, the horizon has been displaced relative to the Poincare
coordinate patch. To understand the relation between the two coordinate patches, we
examine how the bifurcation surface at ρ = L (and any finite value of τ) approaches
the AdS boundary. Combining eqs. (3.7) and (3.16), we find
y−1 + yd =L2
z= eβ L coshu , y2
1 + · · ·+ y2d−1 =
L2
z2r2 = L2 sinh2 u . (3.17)
Recall that r2 = (x1)2 + ... + (xd−1)2. Hence we can see that the bifurcation surface
intersects the asymptotic boundary in the Poincare coordinates with
r2 = z2 sinh2 u = e−2β L2 tanh2 u →u→∞
e−2βL2 . (3.18)
That is, on a sphere of radius R = e−βL in the flat boundary metric. With a bit of
work, one can determine the precise relation between the two coordinate systems on
the boundary of AdSd+1. In fact, one finds precisely the coordinate transformation
(2.25) with R = e−βL and τ = e−βτ .
Interpreting the bulk metric (3.11) as a topological black hole, the boundary CFT
on R ×Hd−1 is in a thermal state. Since the metric is unchanged, the Hawking tem-
perature of the black hole will be T = 1/(2πL), precisely as before. We denote this
temperature as T since it is determined for energies conjugate to time translations in
τ . To compare to the discussion in section 2, we wish to determine the temperature T
conjugate to the time coordinate τ in the metric (2.37). As noted above, these two co-
ordinates are related by the scaling τ = e−βτ . Hence the temperatures are also related
by T = e−βT which yields
T =1
2π e−βL=
1
2πR. (3.19)
Hence we have again reproduced precisely the temperature emerging in our general
considerations of CFT’s in section 2.
Next we turn to the calculation of the horizon entropy, which we reduced to eq. (3.4)
in the introductory remarks above. The expression there is simply proportional to the
area of the horizon, which has the geometry Hd−1. As already noted, the total horizon
– 21 –
entropy diverges but this is precisely the expected result from section 2.2. Both the
boundary and bulk description yield a uniform constant for the entropy density and so
when integrated over the entire Hd−1 geometry, it produces a divergent total entropy.
As discussed in section 2.2, equality of this thermal entropy and the entanglement
entropy requires a certain relation (2.33) between the long distance cut-off introduced
to regulate the thermal entropy in R × Hd−1 and the short distance cut-off required
to regulated the entanglement entropy across the sphere in Minkowski space. We now
show that the same relation naturally appears in the holographic framework. In the
Poincare coordinates (3.9), a short distance cut-off δ in the CFT is implemented by
introducing a minimal radial coordinate zmin which cuts off the asymptotic region of
the AdS geometry. The standard holographic dictionary for these two cut-offs is simply
zmin = δ. On the horizon (in particular, on the bifurcation surface), eq. (3.17) gave the
relation coshu = e−βL/z and so we find the maximum radius to regulate the calculation
of the horizon entropy is naturally given by
coshumax =e−βL
zmin=R
δ. (3.20)
This relation can also be re-written as
exp(−umax) =R
δ−√R2
δ2− 1 ' δ
2R. (3.21)
Hence, to leading order, we find agreement with the CFT expression in eq. (2.33).
We are now prepared to evaluate the horizon entropy (3.4). To facilitate a com-
parison with the results of [13, 14], we first make a change in the radial coordinate on
the hyperbolic plane to x = sinhu. Hence the IR regulator (3.20) becomes
xmax = sinhumax =
√R2
δ2− 1 (3.22)
=R
δ
(1 +
1
2
δ2
R2+ · · ·
).
Given the bulk metric (3.11), the horizon entropy (3.4) becomes
S =2 Γ(d/2)
πd/2−1a∗d Ωd−2
∫ xmax
0
xd−2 dx√1 + x2
, (3.23)
where Ωd−2 = 2π(d−1)/2/Γ((d − 1)/2) is the area of a unit (d − 2)-sphere. This result
precisely matches eq. (4.5) in [14] (up to the replacement x→ ρ).
Hence as in [14], we observe that the leading contribution arising from eq. (3.23)
can be written as
S ' 2π
πd/2Γ(d/2)
d− 2a∗dAd−2
δd−2+ · · · , (3.24)
– 22 –
where Ad−2 = Ωd−2Rd−2 is the ‘area’ of the entangling surface, i.e., an Sd−2 of radius
R. Hence this leading divergence takes precisely the form expected for the ‘area law’
contribution to the entanglement entropy in a d-dimensional CFT [5, 10]. Further we
note that the hyperbolic geometry of the horizon was essential to ensure the leading
power was 1/δd−2 here despite the area integral being (d−1)-dimensional in eq. (3.23).
This divergent contribution to the entanglement entropy is not universal, e.g., see
[5, 10]. However, a universal contribution can be extracted from the subleading terms.
The form of the universal contribution to the entanglement entropy depends on whether
d is odd or even. For even d, the universal term is logarithmic in the cut-off while for odd
d, it is simply a constant term [5, 10]. In the present case, expanding the entanglement
entropy (3.23) in powers of R/δ, we find the following universal contributions:
Suniv =
(−)
d2−1 4 a∗d log(2R/δ) for even d ,
(−)d−12 2π a∗d for odd d .
(3.25)
Further, recall that for even d, a∗d = A, the coefficient of the A-type trace anomaly in
the boundary CFT.
3.2 Entanglement entropy in a cylindrical background
We now apply the results of our general CFT story discussion in section 2.3 to a holo-
graphic calculation of entanglement entropy in an R×Sd−1 background. In particular,
we choose the entangling surface to be (d–2)-dimensional sphere of a fixed angular ra-
dius θ0. We begin, as before, with the description of AdSd+1 as a hyperboloid (3.5)
embedded in R2,d. In this case, we relate the embedding coordinates to the standard
global coordinates on the AdS geometry
y−1 = % cos(t/L) , y0 = % sin(t/L) , yd = % cos θ , (3.26)
y1 = % sin θ cosφ1 , y2 = % sin θ sinφ1 cosφ2 , · · · yd−1 = % sin θ sinφ1 sinφ2 · · · sinφd−2 .
In this case, the constraint (3.5) yields %2 = %2 + L2 and the induced metric becomes
ds2 =d%2
%2
L2 + 1−(%2
L2+ 1
)dt2 + %2
(dθ2 + sin2 θ dΩ2
d−2
). (3.27)
If as usual, we consider the asymptotic limit %→∞ and remove a factor of %2/L2, the
boundary metric reduces to
ds2CFT = −dt2 + L2
(dθ2 + sin2 θ dΩ2
d−2
). (3.28)
That is, we are studying the boundary CFT in the background geometry R × Sd−1,
where the radius of curvature of the sphere is L.
– 23 –
Now we wish to connect this coordinate choice to that giving the hyperbolic foli-
ation in eqs. (3.15) and (3.16). For simplicity of the presentation, we again focus on
the bifurcation surface in the corresponding metric (3.11). We approach the latter by
taking ρ→ L with τ fixed, which yields y0 = 0 and in turn we find t = 0 in eq. (3.26).
Then using eq. (3.16), we find that
ydy−1
=%√
%2 + L2cos θ = tanh β . (3.29)
Hence in the asymptotic limit % → ∞, we see that the bifurcation surface reaches an
angular size on the Sd−1 with
cos θ0 = tanh β or sin θ0 =1
cosh β. (3.30)
This sphere is then the entangling surface in our calculation of entanglement entropy
in the boundary CFT. We can further construct the full relation between the two
coordinate systems and the result is identical to that given in eq. (2.35) with the
substitutions: R = L and τ = τ .8
The entropy is again given by interpreting the hyperbolic foliation (3.11) as a
topological black hole and evaluating the horizon entropy, as in eq. (3.4). However,
we must first determine the appropriate IR regulator to introduce in the integral over
the horizon. Given the global coordinates (3.27), the standard AdS/CFT dictionary
introduces a short distance cut-off in the boundary CFT as a maximum radius: %max =
L2/δ. Using the two expressions for y−1 in the two coordinate systems, we find that
the bifurcation surface intersects this regulator surface at
coshumax =L sin θ0
δ
(1 +
δ2
L2
)1/2
. (3.31)
We can relate this result to the CFT discussion in section 2.3 as follows. There a small
angle δθ was chosen to regulate the calculation of the entanglement entropy. This small
angle defines a small proper distance on the Sd−1 and so we relate this angular regulator
to the short distance regulator above with δ = Lδθ. Then a bit of algebra allows us to
re-write eq. (3.31) as
exp(−umax) 'δθ
2 sin θ0
+ · · · . (3.32)
Hence the holographic framework reproduces (to leading order) the relation (2.39)
between the UV and IR regulators determined purely in our general discussion of
CFT’s.
8Again, we originally derived the transformation (2.35) with this approach.
– 24 –
Our goal is to evaluate the holographic entanglement entropy across the entan-
gling sphere at θ = θ0 which is again found by calculating the horizon entropy of the
topological black hole (3.11). Hence we return to eq. (3.4) which we write as
S =2 Γ(d/2)
πd/2−1a∗d Ωd−2
∫ xmax
0
xd−2 dx√1 + x2
, (3.33)
where, as above, we have chosen x = sinhu as an alternate radial coordinate on the
hyperbolic plane Hd−1. With eq. (3.31), the regulator radius then becomes
xmax = sinhumax =L sin θ0
δ
(1− δ2
L2cot2 θ0
)1/2
. (3.34)
With this result, we observe that the leading contribution arising from eq. (3.24) re-
produces the expected ‘area law’ contribution to the entanglement entropy [5, 10]. The
universal contributions, appearing in the subleading terms [5, 10], now take the form:
Suniv =
(−)
d2−1 4 a∗d log
(2Lδ
sin θ0
)for even d ,
(−)d−12 2π a∗d for odd d .
(3.35)
Recall that here L denotes the radius of curvature of the Sd−1.
This derivation of the holographic entanglement entropy extends the calculations
presented in [13, 14] which only examined the special case θ0 = π/2, i.e., the entangling
sphere was chosen to be the equator of the background Sd−1 in the cylindrical back-
ground. Of course, setting θ0 = π/2 above, we recover the results derived in [13, 14].
Note that this agreement in trivial for odd d, since the universal term is independent
of θ0 (as well as L). Further we note that the above expressions are essentially the
same (up to L → R) as those in eq. (3.25) where the background geometry was just
flat Minkowski space.
4. A CFT calculation
In the previous section, we have found for a broad range of geometries that the universal
contribution to the entanglement entropy for general holographic CFT’s is controlled
by the central charge a∗d. For even dimensional CFT’s, this charge coincides precisely
with the coefficient of the A-type trace anomaly. Much of the motivation for this work
came from [13, 14], where a similar result was found for the entanglement entropy
between the two halves of the sphere in the background geometry R × Sd−1. In fact,
in [14], it was shown that their result actually applied for any CFT in even d, without
reference to holography. Here we would like to show that our present results can also
be extended to general CFT’s.
– 25 –
4.1 Mapping to de Sitter space
In order to calculate the entanglement entropy of the sphere for a CFT, we find it
convenient to use a mapping to (the static patch of) de Sitter space, instead of the
mapping to R × Hd−1. We start again with the flat space metric for d-dimensional
Minkowski space in polar coordinates:
ds2 = −dt2 + dr2 + r2 dΩ2d−2 . (4.1)
Now with the coordinate transformation
t = Rcos θ sinh(τ/R)
1 + cos θ cosh(τ/R), (4.2)
r = Rsin θ
1 + cos θ cosh(τ/R),
we can readily verify the flat space metric (4.1) becomes
ds2 = Ω2[− cos2θ dτ 2 +R2
(dθ2 + sin2θ dΩ2
d−2
)](4.3)
where Ω = (1 + cos θ cosh(τ/R))−1 .
After eliminating the conformal factor Ω2, the remaining metric corresponds to the
static patch of d-dimensional de Sitter space with curvature scale R. The latter identi-
fication may be clearer if we transform to r = R sin θ, which puts the above metric in
the form
ds2 = −(
1− r2
R2
)dτ 2 +
dr2
1− r2
R2
+ r2 dΩ2d−2 . (4.4)
With eq. (4.2), we observe that
τ → ±∞ : (t, r)→ (±R, 0) (4.5)
θ → π
2: (t, r)→ (0, R)
Note that θ = π/2 corresponds to the cosmological horizon at the boundary of the static
patch. In any event, with eq. (4.5), we see that the new coordinates cover precisely the
causal development D of the ball r ≤ R on the surface t = 0.
Recall that inside D, the modular transformations act geometrically along the flow
in eq. (2.13). This transformation corresponds through the conformal mapping (4.2)
to the time translation τ → τ + 2πR s in de Sitter space — just as happened with
the mapping to R × Hd−1. Therefore the modular transformations act geometrically
as time translations in the static patch and the state in the de Sitter geometry is
thermal at temperature T = 1/(2πR) with respect to the Hamiltonian Hτ generating τ
translations, i.e., the density matrix is given by ρ ∼ exp [−2πRHτ ]. Again, this result
generalizes observations made in ref. [16], which examined conformal mappings of free
conformal field theories in d = 4.
– 26 –
4.2 Thermodynamic entropy
As in section 2, the entanglement entropy for the sphere of radius R in flat space is
equivalent to the thermodynamic entropy of the thermal state in de Sitter space. We
are going to use standard thermodynamics in order to compute this entropy. Normal-
izing the thermal density matrix ρ = e−βHτ/tr(e−βHτ ), we calculate the von Neumann
entropy
S = −tr(ρ log ρ) = β tr(ρHτ ) + log tr(e−βHτ )
= βE − W (4.6)
whereW = − logZ denotes the ‘free energy’ of the partition function Z = tr (exp[−2πRHτ ]).
The energy term in (4.6) is, of course, the expectation value of the operator which
generates τ translations. Since the latter translations correspond to a Killing symmetry
of the static patch (4.3), E is just the Killing energy which can be expressed as
E =
∫V
dd−1x√h 〈Tµν 〉 ξµ nν = −
∫V
dd−1x√−g 〈T τ τ 〉 , (4.7)
where the integral runs over V , a constant τ slice out to θ = π/2. Further nµ∂µ ≡√|gττ | ∂τ is the unit vector normal to V and ξµ∂µ ≡ ∂τ is the time translation Killing
vector.
Recall for the present investigation, the quantum field theory can be any general
CFT. Then the state, coming from a conformal transformation of the Minkowski vac-
uum, is invariant under de Sitter symmetry group [36, 37] and we have
〈T µν 〉 = κ δµν , (4.8)
where κ is some constant. Hence the expectation value of the stress tensor is completely
determined by the conformal anomaly (1.2). Note that in eq. (1.2), all of the the Weyl
invariants vanish in de Sitter space, i.e., In = 0, while the Euler density Ed yields a
constant depending on the de Sitter radius R. Hence we can fix the constant κ in
eq. (4.8) as
〈T µν 〉 = −2 (−)d/2AEddδµν , (4.9)
for d even. The energy (4.7) then becomes
E = 2 (−)d/2AEddRd−1Ωd−2
∫ π/2
0
dθ cos θ sind−2 θ . (4.10)
Hence the energy is finite. However, we are only interested in the universal coefficient
of the logarithmic term in the entropy and hence, given this result, we may discard the
– 27 –
energy contribution in eq. (4.6). Note that for d odd, the trace anomaly vanishes and
so we would find E = 0.
It remains to compute the contribution W = − log tr (exp[−2πRHτ ]) in eq. (4.6).
This can be done as usual passing to imaginary time τE and compactifying the Euclidean
time with a period β = 2πR. The metric becomes
ds2 = cos2θ dτ 2 +R2(dθ2 + sin2θ dΩ2
d−2
). (4.11)
This Euclidean manifold is precisely a d-dimensional sphere with radius of curvature
R.9 Note that the periodicity ∆τ = 2πR is precisely that required to avoid a conical
singularity at θ = π/2. Thus, one ends up with the Euclidean path integral on Sd.
Recall that we only need to determine the coefficient of the logarithmic term in the
entropy for even d. Now the free energy has a general expansion
W = − logZ = (non-universal terms) + ad+1 log δ + (finite terms) , (4.12)
where δ is our short distance cut-off and the non-universal terms diverge as inverse
powers of δ. The coefficient ad+1 for a conformal field theory is determined by the
integrated conformal anomaly [39] — for free fields, it is one of the coefficients in the
heat kernel expansion. In order to see this, consider an infinitesimal rescaling of the
metric gµν → (1− 2δλ)gµν . Since10
2√g
δW
δgµν= 〈Tµν 〉+ (divergent terms) , (4.13)
in terms of the renormalized stress tensor 〈Tµν 〉, we have
δW
δλ= −
∫ddx√g 〈T µµ 〉+ (divergent terms) , (4.14)
which is the integrated trace anomaly. On the other hand, due to the conformal invari-
ance of the action, scaling the metric as above must give the same result as keeping
the metric constant but shifting the UV regulator: δ → (1 − δλ)δ. Combining these
expressions with eq. (4.12), one finds
ad+1 =
∫ddx√g 〈T µµ 〉 . (4.15)
9The fact that this metric (4.11) corresponds to the sphere may be more evident after the coordinate
transformation [38]: sin θ = sin θ1 sin θ2 and tan(τ/R) = cos θ2 tan θ1, which transforms the metric to
ds2 = R2(dθ21 + sin2 θ21dθ
22 + sin2 θ1 sin2 θ2dΩ2
d−2
).
10Recall we have made the transition to a Euclidean signature here.
– 28 –
Hence we are left to substitute eq. (1.2) for the trace anomaly and integrate over
the Sd. Here we also need to observe that since the sphere is conformally flat, all of the
Weyl invariants In vanish for the sphere, while our convention in eq. (1.2) is that the
integral of the Euler density on Sd yields 2. Hence for any CFT in even dimensions,
the universal contribution to the entanglement entropy becomes
Suniv = (−1)d2−14A log(R/δ) , (4.16)
which is essentially the same as our holographic result in eq. (3.25). In particular, we
see that the coefficient of the universal term in the entanglement entropy is proportional
to the central charge A. Note that this result (4.16) and eq. (3.25) do not quite agree
on the argument of the logarithm. In eq. (4.16), R was simply inserted as the only
available scale in the problem whereas the result in eq. (3.25) emerged from a detailed
evaluation of the entropy. Hence the mismatch is no surprise. However, the difference
between the two expressions can be simply regarded as a non-universal constant term.
Although we do not present the details here, it is straightforward to extend our
analysis above to the entanglement entropy across a sphere for CFT’s in a cylindrical
background R×Sd−1. The result for the universal contribution for even d again matches
our holographic result (3.35). In particular, the coefficient is controlled by the coefficient
of the A-type trace anomaly and in fact, it is identical to that just above in eq. (4.16).
As above, the present approach would not naturally reveal precisely the same scale over
δ in the argument of the logarithm, as we found in eq. (3.35).
In the odd dimensional case, we have seen E = 0 and so eq. (4.6) reduces to
S = logZ. That is, the entanglement entropy is simply minus the free energy on
a sphere. This case for odd dimensions has recently been examined in [22] for the
particular case of a free scalar field and the results there are in agreement with this
identification.
4.3 Hyperbolic mapping
It is interesting to go through a similar CFT analysis for our mapping (2.25) from flat
space to R × Hd−1. As discussed in section 2.2, in the latter hyperbolic space, we
have a thermal state with T = 1/(2πR). Now, however, this space is not maximally
symmetric and so by symmetry, the stress tensor is only restricted to have a form
T µν = diag(−E , p, · · · , p) , (4.17)
with E and p constants. The background geometry is again conformally flat and so the
Weyl invariants In vanish. Further, the background is the direct product of two lower
dimensional geometries which dictates that the Euler density is also zero. Hence the
– 29 –
trace anomaly (1.2) vanishes in this particular background and in eq. (4.17), we must
have
E = (d− 1) p . (4.18)
To proceed further we must focus on d = 4 since in general, we do not know the value
of the energy density. However, for d = 4, we have the Bunch-Davies-Brown-Cassidy
formula [40] which relates the stress tensor for a given state in flat space to that in
another space obtained by a conformal mapping from flat space. With this expression,
we obtain
E =3a+ c
8π2R4. (4.19)
Here, we have adopted the standard notation for the central charges in four-dimensional
CFT’s, i.e., comparing to eq. (1.2), a = A and c = 16π2B1 [13].11
Now, in contrast to the discussion in the previous section, the energy contribution
in eq. (4.6) is divergent due to the infinite volume of the Hd−1. Using eq. (4.19) for
d = 4, we find a contribution to the logarithmic term in the entropy
2πRE = 8π2R4 E∫ umax
0
du sinh2(u) (4.20)
= (non-universal terms) +1
2(3a+ c) log δ + (finite terms) .
The final result above relies on choosing the maximum radius umax as in eq. (2.33).
The second contribution to the logarithmic term in the entropy (4.6) comes from
the free energy W which should again be determined by the conformal anomaly, as in
eqs. (4.12) and (4.15). However, as noted above, 〈T µµ 〉 = 0 in the present case and so
we have a zero ‘bulk’ contribution from the anomaly. However, implicitly, the manifold
comes with a boundary, i.e., u = umax in the cut-off manifold. Hence the conformal
anomaly should pick up boundary contributions there but unfortunately, at present,
the expression for these terms is unknown. One could take the results for the entropy
(4.16) and the energy (4.20) in d = 4 to determine this boundary contribution as
W = (non-universal terms) +1
2(−5a+ c) log δ . (4.21)
In particular, the term proportional to c in eq. (4.20) must cancel with a boundary
term. It is intriguing that this approach may imply that there are further restrictions
for the boundary conditions that should be used to define the cut-off at u = umax.
11Our conventions are such that these coefficients are normalized to a = 1/360 and c = 1/120 for a
free conformally coupled massless (real) scalar.
– 30 –
5. Discussion
To summarize our results, we have produced a derivation of holographic entanglement
entropy for certain geometries, namely, with spherical entangling surfaces. The deriva-
tion started by finding an appropriate calculation of the entanglement entropy in the
boundary CFT in section 2. Here, we have avoided the usual approach of using the
replica trick [2, 12]. Rather we used conformal transformations to relate the entan-
glement entropy across a spherical entangling surface to the thermal entropy in a new
background geometry, R×Hd−1. While this construction applies for any CFT, it is not
particularly useful in general as it simply relates two difficult problems to one another.
However, in the case of a holographic CFT, the AdS/CFT correspondence translates
the second problem to the question of determining the horizon entropy of a topological
black hole, as described in section 3. The latter is a straightforward calculation using
Wald’s entropy formula (3.2). Hence we have derived an expression for the holographic
entanglement entropy (3.4), which applies for any bulk gravitational theory, albeit for
the specialized case of a spherical entangling surface. We have explicitly considered the
entanglement entropy for flat space, in section 3.1, or for a cylindrical background, in
section 3.2. This discussion would also straightforwardly extend to a spherical entan-
gling surface in a background R×Hd−1. The key difference, however, would be that we
would not start with the vacuum state in this background, rather we take the thermal
state that is equivalent to the vacuum in Rd by the conformal mappings introduced in
section 2.
The present discussion extends the derivation presented in [14], where the analysis
focused on a special case of the geometries considered here. There the background
geometry was chosen to be R × Sd−1 and the entangling surface was placed on the
equator of the Sd−1. We might comment, however, that the enhanced symmetry of
the latter geometry allowed for an alternate derivation which was based on the replica
trick.
In section 2, both the entanglement and thermal entropy were divergent and their
equality was only guaranteed by imposing a relation between the short-distance cut-
off required to regulate the entanglement entropy and the long-distance cut-off used
to regulate the thermal entropy, as in eqs. (2.33) and (2.39). Hence the conformal
mapping introduced an interesting UV/IR relation between the two states of the CFT,
which is reminiscent of the UV/IR connection found in the AdS/CFT correspondence.
The holographic dictionary also naturally reproduced these relations in section 3. In
this case, the horizon of the topological black hole extended out to the AdS boundary
and the desired relations were determined examining the intersection of the horizon
with the UV regulator surface associated with the original background geometry. The
– 31 –
intersection of these two surfaces was also the key feature which distinguished different
horizons connected to boundary spheres with different sizes. Otherwise the horizons are
‘identical’, in that, these surfaces can all be mapped into one another by an isometry
of the AdSd+1 geometry.
Examining the results in eqs. (3.25) and (3.35), we observe that the universal
contribution to the entanglement entropy is proportional to the central charge, a∗d,
which characterizes the boundary CFT. This charge was introduced in [13, 14] where
it was observed that this charge satisfies a holographic c-theorem12 — a result which
was conjectured to extend to general field theories. Following the considerations of the
holographic principle in [45], it was also argued that a∗d gave a measure of the number
of degrees of freedom in the boundary field theory [14]. Given the present results, we
can identify this charge using entanglement entropy for a broader class of geometries,
in particular, if we wish to examine the c-theorem noted above outside of a holographic
framework. Specifically, with eq. (3.25), a∗d can always be identified using a spherical
entangling surface in flat space.
For the case of an even dimensional boundary theory (i.e., d even), the central
charge is precisely that appearing in the A-type trace anomaly (1.2). In this case, the
universal term in the entanglement entropy is proportional to a logarithm of the cut-off
scale δ. Of course, the latter is balanced by another scale to make the argument of
the logarithm dimensionless. In the flat Minkowski background, this scale is naturally
set by the size of the entangling sphere, as in eq. (3.25), since this is the only scale in
the construction. In our result (3.35) for the Einstein universe background, the other
scale is L sin θ0 (up to a factor of 2, which also appears in the flat space calculation).
This scale can be readily identified as the radius of curvature of the entangling sphere.
Hence, in this sense, the same scale appears in the result for both of our calculations.
We also note that our result (3.35) is a simple generalization of the standard result
for the entanglement entropy of a two-dimensional CFT. Given a sub-system of length
` in a full system of size C (with periodic boundary conditions), the entanglement
entropy is given by[2, 46]
S =c
3log
(C
πδsin
π`
C
), (5.1)
where c is the central charge of the d = 2 CFT.13 Of course, for two dimensions, the
cylindrical background considered in section 3.2 reduces to R× S1. Further our result
(3.35) precisely matches the result given above using the relations: a∗d = A = c/12, L =
C/2π and θ0 = π`/C, which are applicable for d = 2. Our result in eq. (3.35) provides
a natural extension of the above expression for d = 2 to higher (even) dimensions.
12See also recent related results in [41, 42, 43, 44].13In general, one might also expect a non-universal constant term to appear on the right-hand side.
– 32 –
Of course, in section 4, we were able to show that the universal contribution to
the entanglement entropy takes the same form as our holographic result (3.25) for any
CFT in any even number of dimensions. A similar result was proven in [14] where
the entangling surface was the equator of the Sd−1 in the static Einstein universe
background. While we did not present the details, it would be straightforward to
extend the arguments of section 4 to cover this case or a spherical entangling surface
of any angular size. Hence, for even d (without any reference to holography), we
have established that the universal contribution to the entanglement entropy across a
spherical entangling sphere is proportional to A.
The calculation in section 4 begins with a mapping of the causal development D to
the static patch in de Sitter space. This approach is similar to that in [19]. However,
in the latter, the thermodynamical entropy is written as
S =
∫ T= 12πR
0
dE(T, V )
T
∣∣∣∣V=const.
. (5.2)
The main difference14 with our approach is that in eq. (5.2), one needs to know the
energy ‘off shell’, i.e., away from the point T = 1/(2πR). This thermal energy is known
only for the case of free fields and so the results in [19] are only determined for free
fields. In contrast, our ‘on shell’ approach only makes reference to the energy and free
energy at T = 1/(2πR) which naturally emerges from the conformal transformation of
the Minkowski vacuum. This essential difference allowed us to derive a general result
for any CFT in any even number of dimensions. However, we might add that it seems
a knowledge of the whole function E(T, V ) would certainly be necessary for computing
the Renyi entropies Sn = (1− n)−1 log(trρn) for general n.
The general connection of the universal terms in entanglement entropy and the
trace anomaly was first noted in [46] for d = 2 and extended to higher dimensions in
[5, 17]. In particular, [17] established that the entanglement entropy for a spherical
entangling surface would be proportional to A in four dimensions. However, we must
comment that the arguments presented in [5, 17] are only completely justified when
there is a rotational symmetry in the transverse space around the entangling surface.
For configurations without this symmetry, additional correction terms must be added to
the entanglement entropy, however, they still seem to have the same general character,
i.e., they are geometric expressions evaluated on the entangling surface with coefficients
linear in the central charges [11, 17, 47].
14This approach must also assume that the entropy at zero temperature vanishes or at least the
relevant logarithmic contribution vanishes. This assumption may be related to the mismatch in [19],
where the coefficient to the logarithmic term is not given by the A-type anomaly for a vector field.
– 33 –
We note that the coefficient of the universal contribution is identical for any sphere
of any size in flat space or in a cylindrical background. Given that all of these ge-
ometries are related by conformal mappings, this reflects the fact that this coefficient
is conformally invariant, as noted in [20], and clarified in the present paper. Another
case then, which also refers to a spherical entangling surface was recently discussed in
[21]. This case is the near-horizon geometry of an extreme black hole, which has a
geometry H2 × Sd−2 and by explicit calculation for a free conformally coupled scalar
in any even dimension, it was shown that the coefficient of the log term was controlled
by A. The event horizon in this geometry can be conformally mapped to a sphere in
flat space. Hence one expects the coefficient for the logarithmic correction to the black
hole entropy is also controlled by the A-type trace anomaly for even d. In fact, given
the conformal invariance of this coefficient, the discussion in [14] would be sufficient
to indicate that A also appears as the coefficient of the universal contribution for any
CFT in geometries considered in section4.
Our calculations of holographic entanglement entropy also yielded results for odd
d in eqs. (3.25) and (3.35). In this case, following [5], we have identified the universal
contribution as the constant term appearing in the expansion in powers of the cut-
off. Hence the result is completely independent of the size of the entangling sphere
or the background geometry. The universality of this constant contribution to the
entanglement entropy is established for a variety of three-dimensional systems [1, 3].
However, we should note that one may worry that in general the precise value of this
constant will depend on the details of the regulator — as discussed in [14].15 We expect
that these issues can be circumvented by considering an appropriate construction with
mutual information — e.g., see [48, 49]. When calculated with two separate regions,
the latter is free of divergences and any regulator ambiguities.
Recall our result in section 4 for the case of odd dimensional CFT’s. Namely, the
entanglement entropy for a spherical entangling surface Sd−2 is precisely minus the free
energy of the CFT on a sphere Sd, i.e., we have
S = logZ for odd d . (5.3)
This result can be related to a calculation of entanglement entropy described in [14].
There, the initial problem was to determine the entanglement entropy of a d-dimensional
CFT on R × Sd−1 when the entangling surface Σ was chosen to be the equator of the
sphere. The approach taken was to apply the geometric approach to the replica trick
[12], where one evaluates the partition function on the background geometry with an
infinitesimal conical defect at Σ. However, this procedure is only well-defined if there is
15RCM thanks M. Smolkin, A. Schwimmer and S. Theisen for discussions on this point.
– 34 –
a rotational symmetry about this surface and so in [14], the R× Sd−1 background was
mapped to Sd using a conformal transformation. The expression for the entanglement
entropy then becomes
S = limε→0
(∂
∂ε+ 1
)logZ1−ε . (5.4)
where the partition function is evaluated on a ‘(1− ε)-fold cover’ of the d-dimensional
sphere, with an infinitesimal conical defect ∆θ = 2π(1− ε) at Σ. In [14], this construc-
tion was applied to determine the universal contribution to the entanglement entropy
for even d. However, it applies just as well for the case of odd d. Hence comparing to
eqs. (5.3) and (5.4), we see that the leading variation of the sphere partition function
must vanish in the latter equation, i.e., ∂εZ|ε=0 = 0. This vanishing of the variation of
Z is analogous to the vanishing of the energy contribution in eq. (4.6) for odd d.
We should note that the expressions on both sides of eq. (5.3) are expected to
diverge and so, as stressed in section 2, care must be taken in applying a consistent
regulator to ensure the equality of the two quantities. As noted before, this equality
was established for free conformal scalar fields in odd dimensions by [22]. There, in
fact, the heat kernel regulator completely eliminated the divergences on both sides of
eq. (5.3). With a general regulator, where divergent terms still appear, one still expects
that S and logZ will be equal order by order in the cut-off scale. In particular, the
universal constant contribution to the entanglement entropy must match the constant
term in free energy.
This last observation is of interest in connection to a recent discussion of N = 2
superconformal field theories in three dimensions in [50]. There the author provides
evidence that, for these theories, the sphere partition function plays a very similar role
to the central charge a in four-dimensional theories with N = 1 supersymmetry. In
particular, as function of possible trial R-charges, (the finite part of) Z is extremized by
the exact superconformal R-charge, in analogy to a-maximization in four dimensions
[51]. Further, given the connection between a-maximization and the c-theorem for the
corresponding field theories [52], one is naturally lead to speculate that Z-maximization
may provide a framework to develop the analog of c-theorem for supersymmetric the-
ories in three dimensions [50]. Now, the identity in eq. (5.3) connects this suggestion
to the broader conjecture of [13, 14]. Motivated by holographic evidence, the authors
there proposed that the central charge a∗d, which appears in the universal contribution
to the entanglement entropy, should evolve monotonically under RG flows. Focussing
on three dimensions, eq. (5.3) would yield
logZ|finite = Suniv = −2π a∗3 . (5.5)
– 35 –
Hence the holographic results of [13, 14] indicate that any d = 3 supersymmetric
gauge theories with a gravity dual will satisfy the desired c-theorem. Alternatively,
if the field theoretic approach of [50] can be extended to establish a c-theorem as a
consequence of Z-maximization, this would provide further evidence for the general
c-theorem conjectured in [13, 14] for quantum field theories in any (odd) number of
spacetime dimensions
Given our new results, it is interesting to consider some explicit applications, in
particular, to consider higher curvature corrections in the holographic entanglement
entropy in various string models. A well-known set of corrections appear at order
(curvature)4 in all superstring theories [53]. The supersymmetric completion of this
term in type IIb string theory was used to explicitly construct all of the interactions
involving the curvature and the Ramond-Ramond five-form [54]. These interactions
are all naturally written in terms of the Weyl tensor and certain tensors constructed
from the five-form. Now in a holographic setting, we are considering a supersymmetric
reduction of the form AdS5 ×M5 and in such a background, both the Weyl tensor
and the relevant tensors for the five-form vanish. Hence these interactions do not
modify the background geometry nor do they contribute to the Wald entropy (3.2) of
the topological black hole. That is, these particular higher order corrections to the
superstring action leave the entanglement entropy unchanged. Note that implicitly we
have extended the discussion to the full ten dimensions of the superstring theory and
so the horizon has the geometry H3 ×M5. We could first reduce the ten-dimensional
theory to five dimensions, following [55], and present the argument entirely in the
context of AdS5 as in the main text. Of course, the result is unchanged.
Further, this result, namely that the entanglement entropy was unchanged, should
have been expected. These R4 interactions appear at order α′3 with a tree-level and
one-loop (i.e., g2s) contribution. From the perspective of the boundary theory then,
these terms will introduce corrections of order 1/λ3/2 and λ1/2/N2c [56]. In particular
then, these depend on the ’t Hooft coupling λ. However, our analysis indicates that
the universal contribution to the entanglement entropy should be proportion to the
central charge A (which is commonly denoted a in four dimensions). Further in a
superconformal gauge theory, it is known that the central charges are independent
of the gauge coupling [57]. Hence, this universal contribution should not receive any
corrections depending on the ’t Hooft coupling, which is in accord with our gravity
calculations. We might add that for N = 4 super-Yang-Mills theory in the limit of
zero coupling (i.e., the free field limit), numerical calculations [58] of the entanglement
entropy for a spherical entangling surface in flat space explicitly confirm that the results
match the strong coupling result. Hence these calculations also confirm the same
independence of the gauge coupling.
– 36 –
There are other interesting string theory models where curvature-squared terms
arise in a holographic context [59]. These terms originate from the presence of D-branes
in the construction of these backgrounds [60]. We do not present the details here but
we comment that the presence of these terms does effect the two central charges, a and
c, of the dual CFT. In particular, the difference c− a is controlled by the coefficient of
this higher curvature interaction. However, we are again discussing these terms in the
context where they appear in a controlled perturbative expansion in string theory and
hence they can be modified by field redefinitions. In particular then, if this interaction
is written as RabcdRabcd, it contributes by modifying both the background curvature
of the AdS5 and the expression for the central charge A. Of course, this term also
contributes to the Wald entropy and the modifications are consistent with are final
result (3.25) and (3.35) where A appears in the coefficient of the universal contribution
to the entanglement entropy. However, with field redefinitions, the interaction can also
be written as CabcdCabcd, in which case, neither the background curvature nor the central
charge a are modified. Further, this term will not contribute to the Wald entropy and
so again the results are consistent with our expressions for the entanglement entropy.
One issue which our discussion highlights is the close connection between holo-
graphic entanglement entropy and black hole entropy. Indeed an eternal AdS black
hole contains two asymptotically AdS regions and it has been argued that in this case
the horizon entropy corresponds to the entanglement entropy between the CFT on one
boundary and its thermofield double on the other boundary [61]. Our construction
essentially uses this interpretation for the topological black hole which corresponds to
a pure AdS spacetime. The key difference from the usual interpretation is that the two
asymptotically AdS regions are complementary portions of the same AdS geometry. It
would be interesting to understand if a similar interpretation is possible for topological
black holes which are endowed with charge charge and/or rotation [62]. Another useful
direction would be to investigate whether AdS space can be foliated in other ways to
produce ‘topological black holes’ with different horizon geometries. These may then
form the basis of a derivation of the holographic entanglement entropy for entangling
surfaces with new geometries.
Of course, our derivation puts the standard proposal of [4, 5] on a firmer footing
since we find agreement with eq. (1.1) when the bulk theory is just Einstein gravity.
We might note that the present calculations do not seem to involve the extremization
of some functional over a family of bulk surfaces. However, given our results, a natural
guess might be that when the bulk gravity theory includes higher curvature terms, we
should extend the definition of holographic entanglement entropy (1.1) to extremize the
Wald entropy (3.2) evaluated on bulk surfaces homologous to the boundary region of
interest. Unfortunately, one can easily show that this procedure does not produce the
– 37 –
correct entanglement entropy in general, however, interesting progress has still been
made for certain classes of higher curvature theories [11, 41]. At the same time, we can
add that for generic entangling surfaces, the bulk surface determining the holographic
entanglement entropy will not play the role of the event horizon for some black hole
[11, 41].
Again, our derivation of holographic entanglement entropy discards the replica trick
and instead we are relying on invariance of the entanglement entropy of the bound-
ary CFT under conformal mappings. Of course, it would be interesting to extend
our approach to produce a derivation for more general geometries, i.e., non-spherical
entangling surfaces in different background spacetimes. A common feature of the con-
formal mappings in sections 3.1 and 3.2, which seems important, is that the entangling
sphere is mapped to space-like infinity in R × Hd−1. While similar transformations
mapping the entangling surface to infinity are easily constructed for other geometries,
e.g., Sd−2−n × Rn, the resulting background is typically time-dependent and it is not
evident what the state of the CFT is in the new background. Hence if further progress
is to be made with this approach, additional insights will be needed with respect to the
most useful conformal mapping to apply for a given geometry. It may be instructive
to translate the discussion of section 4 to a holographic derivation of the entanglement
entropy of a spherical entangling surface. In any event, there remain many interesting
questions to explore with regards to holographic entanglement entropy.
Acknowledgments: RCM thanks Stuart Dowker, Ben Freivogel, Janet Hung, Alex
Maloney, Aninda Sinha, Misha Smolkin, and Lenny Susskind for useful discussions.
HC is grateful to the Perimeter Institute for hospitality during the initial stages of
this work. Research at Perimeter Institute is supported by the Government of Canada
through Industry Canada and by the Province of Ontario through the Ministry of
Research & Innovation. RCM also acknowledges support from an NSERC Discovery
grant and funding from the Canadian Institute for Advanced Research. MH and HC
acknowledge support from CONICET and Universidad Nacional de Cuyo, Argentina.
References
[1] See, for example:
M. Levin and X. G. Wen, “Detecting Topological Order in a Ground State Wave
Function,” Phys. Rev. Lett. 96, 110405 (2006) [arXiv:cond-mat/0510613];
– 38 –
A. Kitaev and J. Preskill, “Topological entanglement entropy,” Phys. Rev. Lett. 96,
110404 (2006) [arXiv:hep-th/0510092].
[2] See, for example:
P. Calabrese and J. L. Cardy, “Entanglement entropy and quantum field theory,” J.
Stat. Mech. 0406, P002 (2004) [arXiv:hep-th/0405152];
P. Calabrese and J. L. Cardy, “Entanglement entropy and quantum field theory: A
non-technical introduction,” Int. J. Quant. Inf. 4, 429 (2006) [arXiv:quant-ph/0505193].
[3] B. Hsu, M. Mulligan, E. Fradkin and E.A. Kim, “Universal entanglement entropy in 2D
conformal quantum critical points,” Phys. Rev. B 79, 115421 (2009) [arXiv:0812.0203].
[4] S. Ryu and T. Takayanagi, “Holographic derivation of entanglement entropy from
AdS/CFT,” Phys. Rev. Lett. 96, 181602 (2006) [arXiv:hep-th/0603001].
[5] S. Ryu and T. Takayanagi, “Aspects of holographic entanglement entropy,” JHEP
0608, 045 (2006) [arXiv:hep-th/0605073].
[6] I. R. Klebanov, D. Kutasov and A. Murugan, “Entanglement as a Probe of
Confinement,” Nucl. Phys. B 796, 274 (2008) [arXiv:0709.2140 [hep-th]].
[7] M. Van Raamsdonk, “Comments on quantum gravity and entanglement,”
arXiv:0907.2939 [hep-th];
M. Van Raamsdonk, “Building up spacetime with quantum entanglement,” Gen. Rel.
Grav. 42, 2323 (2010) [arXiv:1005.3035 [hep-th]].
[8] M. Headrick, “Entanglement Renyi entropies in holographic theories,” arXiv:1006.0047
[hep-th].
[9] D. V. Fursaev, “Proof of the holographic formula for entanglement entropy,” JHEP
0609, 018 (2006) [arXiv:hep-th/0606184].
[10] T. Nishioka, S. Ryu and T. Takayanagi, “Holographic Entanglement Entropy: An
Overview,” J. Phys. A 42, 504008 (2009) [arXiv:0905.0932 [hep-th]].
[11] L. Y. Hung, R. C. Myers and M. Smolkin, “On Holographic Entanglement Entropy and
Higher Curvature Gravity,” arXiv:1101.5813 [hep-th].
[12] C. G. Callan and F. Wilczek, “On geometric entropy,” Phys. Lett. B 333, 55 (1994)
[arXiv:hep-th/9401072].
[13] R. C. Myers and A. Sinha, “Seeing a c-theorem with holography,” Phys. Rev. D 82,
046006 (2010) [arXiv:1006.1263 [hep-th]].
[14] R. C. Myers and A. Sinha, “Holographic c-theorems in arbitrary dimensions,”
arXiv:1011.5819 [hep-th].
– 39 –
[15] H. Casini and M. Huerta, “Entanglement entropy for the n-sphere,” Phys. Lett. B 694,
167 (2010) [arXiv:1007.1813 [hep-th]].
[16] P. Candelas and J. S. Dowker, “Field Theories On Conformally Related Space-Times:
Some Global Considerations,” Phys. Rev. D 19, 2902 (1979).
[17] S. N. Solodukhin, “Entanglement entropy, conformal invariance and extrinsic
geometry,” Phys. Lett. B 665, 305 (2008) [arXiv:0802.3117 [hep-th]].
[18] See, for example:
M. J. Duff, “Observations on Conformal Anomalies,” Nucl. Phys. B125, 334 (1977);
M. J. Duff, “Twenty years of the Weyl anomaly,” Class. Quant. Grav. 11, 1387-1404
(1994) [hep-th/9308075];
S. Deser and A. Schwimmer, “Geometric classification of conformal anomalies in
arbitrary dimensions,” Phys. Lett. B309, 279-284 (1993) [hep-th/9302047].
[19] J. S. Dowker, “Entanglement entropy for even spheres,” arXiv:1009.3854 [hep-th].
[20] J. S. Dowker, “Hyperspherical entanglement entropy,” J. Phys. A A43, 445402 (2010)
[arXiv:1007.3865 [hep-th]].
[21] S. N. Solodukhin, “Entanglement entropy of round spheres,” Phys. Lett. B693,
605-608 (2010). [arXiv:1008.4314 [hep-th]].
[22] J. S. Dowker, “Entanglement entropy for odd spheres,” [arXiv:1012.1548 [hep-th]].
[23] R. Haag, “Local quantum physics: Fields, particles, algebras”, Berlin, Germany:
Springer (1992) (Texts and monographs in physics).
[24] See, for example:
H. Li and F. D. M. Haldane, “Entanglement Spectrum as a Generalization of
Entanglement Entropy: Identification of Topological Order in Non-Abelian Fractional
Quantum Hall Effect States,” Phys. Rev. Lett. 101, 010504 (2008) [arXiv:0805.0332
[cond-mat.mes-hall]];
P. Calabrese and A. Lefevre, “Entanglement spectrum in one-dimensional systems,”
Phys. Rev. A 78, 032329 (2008);
A. M. Turner, F. Pollmann and E. Berg, “Topological Phases of One-Dimensional
Fermions: An Entanglement Point of View,” Phys. Rev. B 83, 075102 (2011)
[arXiv:1008.4346 [cond-mat.str-el]];
L. Fidkowski, “Entanglement spectrum of topological insulators and superconductors,”
Phys. Rev. Lett. 104, 130502 (2010) [arXiv:0909.2654 [cond-mat.str-el]];
H. Yao and X.-L. Qi, “Entanglement entropy and entanglement spectrum of the Kitaev
model,” Phys. Rev. Lett. 105, 080501 (2010) [arXiv:1001.1165 [cond-mat.str-el]].
– 40 –
[25] J. J. Bisognano and E. H. Wichmann, “On The Duality Condition For Quantum
Fields,” J. Math. Phys. 17, 303 (1976);
J. J. Bisognano and E. H. Wichmann, “On The Duality Condition For A Hermitian
Scalar Field,” J. Math. Phys. 16, 985 (1975).
[26] W. G. Unruh, “Notes on black hole evaporation,” Phys. Rev. D 14, 870 (1976).
[27] P. D. Hislop and R. Longo, “Modular Structure Of The Local Algebras Associated
With The Free Massless Scalar Field Theory,” Commun. Math. Phys. 84, 71 (1982);
[28] H. Casini, M. Huerta, “Remarks on the entanglement entropy for disconnected
regions,” JHEP 0903, 048 (2009). [arXiv:0812.1773 [hep-th]].
[29] R. Emparan, “AdS/CFT duals of topological black holes and the entropy of
zero-energy states,” JHEP 9906, 036 (1999) [arXiv:hep-th/9906040].
[30] See, for example:
S. Aminneborg, I. Bengtsson, S. Holst and P. Peldan, “Making Anti-de Sitter Black
Holes,” Class. Quant. Grav. 13, 2707 (1996) [arXiv:gr-qc/9604005];
D. R. Brill, J. Louko and P. Peldan, “Thermodynamics of (3+1)-dimensional black
holes with toroidal or higher genus horizons,” Phys. Rev. D 56, 3600 (1997)
[arXiv:gr-qc/9705012];
L. Vanzo, “Black holes with unusual topology,” Phys. Rev. D 56, 6475 (1997)
[arXiv:gr-qc/9705004];
R. B. Mann, “Pair production of topological anti-de Sitter black holes,” Class. Quant.
Grav. 14, L109 (1997) [arXiv:gr-qc/9607071];
D. Birmingham, “Topological black holes in anti-de Sitter space,” Class. Quant. Grav.
16, 1197 (1999) [arXiv:hep-th/9808032]. R. Emparan, “AdS membranes wrapped on
surfaces of arbitrary genus,” Phys. Lett. B 432, 74 (1998) [arXiv:hep-th/9804031].
[31] See, for example:
D. J. Gross and J. H. Sloan, “The Quartic Effective Action for the Heterotic String,”
Nucl. Phys. B 291, 41 (1987);
C. G. Callan, E. J. Martinec, M. J. Perry and D. Friedan, “Strings In Background
Fields,” Nucl. Phys. B 262, 593 (1985).
[32] See, for example:
R. C. Myers, M. F. Paulos and A. Sinha, “Holographic studies of quasi-topological
gravity,” JHEP 1008, 035 (2010) [arXiv:1004.2055 [hep-th]];
A. Buchel, J. Escobedo, R. C. Myers, M. F. Paulos, A. Sinha and M. Smolkin,
“Holographic GB gravity in arbitrary dimensions,” JHEP 1003, 111 (2010)
[arXiv:0911.4257 [hep-th]];
– 41 –
J. de Boer, M. Kulaxizi and A. Parnachev, “AdS7/CFT6, Gauss-Bonnet Gravity, and
Viscosity Bound,” JHEP 1003, 087 (2010) [arXiv:0910.5347 [hep-th]];
X. O. Camanho and J. D. Edelstein, “Causality constraints in AdS/CFT from
conformal collider physics and Gauss-Bonnet gravity,” JHEP 1004, 007 (2010)
[arXiv:0911.3160 [hep-th]];
J. de Boer, M. Kulaxizi and A. Parnachev, “Holographic Lovelock Gravities and Black
Holes,” JHEP 1006, 008 (2010) [arXiv:0912.1877 [hep-th]];
X. O. Camanho and J. D. Edelstein, “Causality in AdS/CFT and Lovelock theory,”
JHEP 1006, 099 (2010) [arXiv:0912.1944 [hep-th]];
X. O. Camanho, J. D. Edelstein and M. F. Paulos, “Lovelock theories, holography and
the fate of the viscosity bound,” arXiv:1010.1682 [hep-th].
[33] D. Lovelock, “The Einstein tensor and its generalizations,” J. Math. Phys. 12, 498
(1971);
D. Lovelock, “Divergence-free tensorial concomitants,” Aequationes Math. 4, 127
(1970).
[34] R. C. Myers and B. Robinson, “Black Holes in Quasi-topological Gravity,” JHEP
1008, 067 (2010) [arXiv:1003.5357 [gr-qc]].
[35] R. M. Wald, “Black hole entropy is the Noether charge,” Phys. Rev. D 48, 3427 (1993)
[arXiv:gr-qc/9307038];
T. Jacobson, G. Kang and R. C. Myers, “On Black Hole Entropy,” Phys. Rev. D 49,
6587 (1994) [arXiv:gr-qc/9312023];
V. Iyer and R. M. Wald, “Some properties of Noether charge and a proposal for
dynamical black hole entropy,” Phys. Rev. D 50, 846 (1994) [arXiv:gr-qc/9403028].
[36] N. D. Birrell, P. C. W. Davies, “Quantum Fields In Curved Space,” Cambridge, Uk:
Univ. Pr. ( 1982) 340p.
[37] R. Laflamme, “Geometry And Thermofields,” Nucl. Phys. B324, 233 (1989).
[38] D. V. Fursaev, G. Miele, “Finite temperature scalar field theory in static de Sitter
space,” Phys. Rev. D49, 987-998 (1994). [hep-th/9302078].
[39] See, for example:
D. V. Vassilevich, “Heat kernel expansion: User’s manual,” Phys. Rept. 388, 279
(2003) [arXiv:hep-th/0306138].
[40] L. S. Brown and J. P. Cassidy, Phys. Rev. D 16, 1712 (1977);
T. S. Bunch and P. C. W. Davies, “Quantum Field Theory In De Sitter Space:
Renormalization By Point Splitting,” Proc. Roy. Soc. Lond. A 360, 117 (1978).
– 42 –
[41] J. de Boer, M. Kulaxizi and A. Parnachev, “Holographic Entanglement Entropy in
Lovelock Gravities,” arXiv:1101.5781 [hep-th].
[42] J. T. Liu, W. Sabra and Z. Zhao, “Holographic c-theorems and higher derivative
gravity,” arXiv:1012.3382 [hep-th].
[43] A. Sinha, “On higher derivative gravity, c-theorems and cosmology,” arXiv:1008.4315
[hep-th].
[44] M. F. Paulos, “Holographic phase space: c-functions and black holes as renormalization
group flows,” arXiv:1101.5993 [hep-th].
[45] L. Susskind and E. Witten, “The holographic bound in anti-de Sitter space,”
arXiv:hep-th/9805114.
[46] C. Holzhey, F. Larsen and F. Wilczek, “Geometric and renormalized entropy in
conformal field theory,” Nucl. Phys. B 424, 443 (1994) [arXiv:hep-th/9403108].
[47] A. Schwimmer and S. Theisen, “Entanglement Entropy, Trace Anomalies and
Holography,” Nucl. Phys. B 801, 1 (2008) [arXiv:0802.1017 [hep-th]].
[48] H. Casini, “Mutual information challenges entropy bounds,” Class. Quant. Grav. 24,
1293 (2007) [arXiv:gr-qc/0609126];
H. Casini and M. Huerta, “Remarks on the entanglement entropy for disconnected
regions,” JHEP 0903, 048 (2009) [arXiv:0812.1773 [hep-th]];
H. Casini and M. Huerta, “Entanglement entropy in free quantum field theory,” J.
Phys. A 42, 504007 (2009) [arXiv:0905.2562 [hep-th]].
[49] B. Swingle, “Mutual information and the structure of entanglement in quantum field
theory,” arXiv:1010.4038 [quant-ph].
[50] D. L. Jafferis, “The Exact Superconformal R-Symmetry Extremizes Z,”
arXiv:1012.3210 [hep-th].
[51] K. A. Intriligator and B. Wecht, “The exact superconformal R-symmetry maximizes
a,” Nucl. Phys. B 667, 183 (2003) [arXiv:hep-th/0304128].
[52] E. Barnes, K. A. Intriligator, B. Wecht and J. Wright, “Evidence for the strongest
version of the 4d a-theorem, via a-maximization along RG flows,” Nucl. Phys. B 702,
131 (2004) [arXiv:hep-th/0408156].
[53] D. J. Gross and E. Witten, “Superstring Modifications Of Einstein’s Equations,” Nucl.
Phys. B 277, 1 (1986);
M. T. Grisaru, A. E. M. van de Ven and D. Zanon, “Four Loop Beta Function For The
N=1 And N=2 Supersymmetric Nonlinear Sigma Model In Two-Dimensions,” Phys.
Lett. B 173, 423 (1986).
– 43 –
[54] M. F. Paulos, “Higher derivative terms including the Ramond-Ramond five-form,”
JHEP 0810, 047 (2008) [arXiv:0804.0763 [hep-th]].
[55] A. Buchel, R. C. Myers, M. F. Paulos and A. Sinha, “Universal holographic
hydrodynamics at finite coupling,” Phys. Lett. B 669, 364 (2008) [arXiv:0808.1837
[hep-th]].
[56] See, for example:
M. B. Green and M. Gutperle, “Effects of D-instantons,” Nucl. Phys. B 498, 195
(1997) [arXiv:hep-th/9701093];
R. C. Myers, M. F. Paulos and A. Sinha, “Quantum corrections to eta/s,” Phys. Rev.
D 79, 041901 (2009) [arXiv:0806.2156 [hep-th]].
[57] D. Anselmi, D. Z. Freedman, M. T. Grisaru and A. A. Johansen, “Universality of the
operator product expansions of SCFT(4),” Phys. Lett. B 394, 329 (1997)
[arXiv:hep-th/9608125];
D. Anselmi, D. Z. Freedman, M. T. Grisaru and A. A. Johansen, “Nonperturbative
formulas for central functions of supersymmetric gauge theories,” Nucl. Phys. B 526,
543 (1998) [arXiv:hep-th/9708042].
[58] R. Lohmayer, H. Neuberger, A. Schwimmer and S. Theisen, “Numerical determination
of entanglement entropy for a sphere,” Phys. Lett. B 685, 222 (2010) [arXiv:0911.4283
[hep-lat]].
[59] A. Buchel, R. C. Myers and A. Sinha, “Beyond η/s = 1/4π,” JHEP 0903, 084 (2009)
[arXiv:0812.2521 [hep-th]];
Y. Kats and P. Petrov, “Effect of curvature squared corrections in AdS on the viscosity
of the dual gauge theory,” JHEP 0901, 044 (2009) [arXiv:0712.0743 [hep-th]].
[60] C. P. Bachas, P. Bain and M. B. Green, “Curvature terms in D-brane actions and their
M-theory origin,” JHEP 9905, 011 (1999) [arXiv:hep-th/9903210].
[61] J. M. Maldacena, “Eternal black holes in Anti-de-Sitter,” JHEP 0304, 021 (2003)
[arXiv:hep-th/0106112].
[62] See, for example:
D. Klemm, V. Moretti and L. Vanzo, “Rotating topological black holes,” Phys. Rev. D
57, 6127 (1998) [Erratum-ibid. D 60, 109902 (1999)] [arXiv:gr-qc/9710123];
M. H. Dehghani, “Rotating topological black holes in various dimensions and
AdS/CFT correspondence,” Phys. Rev. D 65, 124002 (2002) [arXiv:hep-th/0203049];
R. B. Mann, “Charged topological black hole pair creation,” Nucl. Phys. B 516, 357
(1998) [arXiv:hep-th/9705223];
R. G. Cai and A. Wang, “Thermodynamics and stability of hyperbolic charged black
holes,” Phys. Rev. D 70, 064013 (2004) [arXiv:hep-th/0406057].
– 44 –
Recommended